Anda di halaman 1dari 9

Production Logging Tool Behavior

in Two-Phase Inclined Flow


A.D. Hill, * SPE, Marathon Oil CO.
T. Oolman, Marathon Oil Co.

Summary
Production logging tools obtained from three service move rapidly along the upper side of the pipe,
companies were tested in one-, two-, and three-phase establishing a circulating current that caused downflow
flow in an inclinable flow test facility. These tests were along the lower side of the pipe. We repeatedly observed
inspired by the often confusing production logs obtained this pattern even though the net flow of all fluids was up-
in deviated wells. The experiments have shown that two- ward. Because of this nonuniform flow profile in
phase flow regimes in inclined pipes differ markedly multiphase flow in inclined wells, responses of produc-
from those in vertical flow at similar flow rates and that tion logging tools that measure localized quantities do
these flow regimes adversely affect production logging not represent volumetric flow rates, average densities, or
tool responses. The experiments performed, the behavior holdups.
of the production logging tools, and a statistical analysis In this type of flow situation, spinner-type velocity
of the data are discussed. tools often indicate reverse flow, especially when they
Introduction are not centralized. Density and capacitance tools tend to
be immersed primarily in the denser phase in a
Production logging tools obtained from three service multiphase flow situation, preventing them from in-
companies were tested in one-, two-, and three-phase dicating average flow properties. Capacitance tools also
flow in an inclined flow test facility at Marathon Oil exhibited nonlinear calibrations in static fluids, which
Co. 's Denver Research Center. Flow rates were chosen makes their response suspect in a flowing stream. Flow
to simulate downhole conditions in producing wells in concentrating flow meters were the best tools tested for
the McArthur River field, where production logging is flow-rate measurements because they force most of the
used extensively. Liquid flow rates ranged from 70 to flow stream through a turbine inside the tools.
3,400 BID (II to 540 m 3 /d) and gas rates ranged from 0 A film made during the testing of production logging
to 7.5 McflD (0 to 210 m 3 /d) during the tests. To tools illustrates the profound effect of inclined,
simulate the wide range of well deviations in this off- multiphase flow on these devices. A 5-minute videotape
shore field, the pipe angle also was varied from vertical version of this film is available from SPE. *
to 60 from vertical. Production logging tool responses We used multivariate regression analysis to correlate
were measured over this range of conditions to under- the responses of two of the tool strings with the flow
stand tool behavior and to try to correlate measurements rates of each fluid and with the deviation angle. These
with actual flow rates. correlations were used to predict the fluid flow rates
Flow patterns observed in multi phase flow in an in- from specific sets of tool response. Correlations ob-
clined pipe were found to affect production logging tool tained, however, were not sufficiently accurate for use in
responses dramatically. In gas/liquid flow in a slanted production logging interpretation. Improvements in tool
pipe, slug flow was the predominant flow regime, design will be necessary before this technique can be ap-
whereas in oil/water two-phase flow, a stratified flow plied to field interpretations. The technique does appear
was common. In both cases, the lighter phase would valid for tools that measure average fluid properties
across the pipe .
Now with U. of Texas.
'Contact SPE Continuing Education Office, 6200 N. Central Expwy., Drawer 64706,
01492136/82/0101-0208$00.25 Dallas, TX 75206. A 3/.-in. (1.9-cm) videotape format is available at $35.00 prepaid
Copyrigh1 1982 Society of Petroleum Engineers of AIME (includes "surface mail" postage). Other formats are available at higher cost.

2432 JOURNAL OF PETROLEUM TECHNOLOGY


Spinner Tool

Fig. 1-lnclined flow test facility.

Inclined Flow Test Facilities


Two flow test facilities have been constructed at our
Denver research center for testing production logging
tools in inclined, multiphase flow. The first facility con-
sisted of a steel pipe loop 20 ft (6.1 m) long with a 6-in.
(152-mm) ID and was inclinable from vertical to
horizontal by means of a winch . Nine-inch (299-mm)
Plexiglas viewports were included in the flow loop
for the observation of flow regimes. The new facility has Fig. 2-Nonuniform flow in gas/liquid two-phase flow.
an inclinable section 40 ft (12.2 m) long, with 8 I h-in.
(216-mm) and 4 Y2-in . (114-mm) ID acrylic plastic pipes flowing stream being water. On some runs , additional
the entire length of the test section . A hydraulic piston is intermediate fractions of water were measured.
used to raise and lower the flow loop. The new flow At each setting of flow rate and angle, the responses
facility is shown in Fig. 1. from whatever tools were in the system, absolute
Separate pumps, flowmeters, and control valves were pressure, temperature, and two t:..p readings were taken.
used for kerosene and water flow, with the two streams On many runs, an observed flow regime also was record-
being combined shortly before they entered the inclined ed . The t:..p measurements were generally a t:..p in the
section. Gas also was added just upstream of the test sec- upflow side of the test section across 12 ft (3.7 m) and a
tion after passing through a rotameter. Oile large tank is t:..p across the tool. The t:..p across the tool was important
used for kerosene/water separation and for storage. for the flow-concentrating-type devices because it in-
dicated how well the tools were capturing the flow
Experimental Procedures stream.
A total of 32 experimental runs was made with the tools Whenever possible, time-averaged digital measure-
from the three service companies. In general, a new run ments were used for the tool responses. Usually, the
number was assigned to an experiment when a major pulses put out by a particular tool would be summed for
change was made that required shutting down the flow 100 seconds, and this sum would be recorded as the tool
system (such as removing the tool from the test section). response. Measurements taken in this manner were much
In each run, the pipe angle was varied from 0 to 60 more reproducible than instantaneous readings from the
from vertical , with readings taken at 0, 5, 15, 30 , 45, tool, which often fluctuated widely in two-phase flow .
and 60 . Measurements were made at each angle for
each setting of the flow rates. Liquid flow rates ranged Flow Regimes Observed
from 170 BID (27 m 3 /d) to 3,400 BID (540 m 3 /d) on Observations of flow regimes during tests of production
most tests, though in some cases flow rates as high as logging tools in inclined pipe quickly indicated their
8,500 BID (1350 m 3 /d) were used . When gas was in- dominant effect on production logging tool responses.
cluded in a run, readings were taken at nominal gas flow The tools tested require a relatively homogeneous flow
rates of 0, 1.5, and 7.5 McflD (0, 42, and 210 m 3 /d). stream (such as bubble flow) to provide measurements of
These rates were chosen to represent the downhole average fluid properties. Bubble flow occurs in vertical
volumetric rate that might be expected in a Cook Inlet upflow under certain flow conditions; these conditions
well having a GOR at the surface in the range of 300 to and other common flow patterns observed in vertical or
400 std cu ft/bbl (54 to 72 std m 31m 3). When kerosene horizontal two-phase flow are summarized by Govier
and water were flowing simultaneously, measurements and Aziz. 1 However, flow regimes observed under most
were generally taken with 0, 10,50,90, and 100 % of the two-phase flow conditions in inclined pipe were far from
OCTOBER 1982 2433
7 r-----------------------------~ most of the pipe volume. Steady reverse flow of water
along the bottom of the pipe was observed as in the
25 gpm Kerosene
25 gpm Water gaslliquid case. Three-phase flow of gas, oil, and water
6 resembled the two-phase gas/liquid case because the tur-
bulence caused by gas slugging appeared to mix the two
liquids substantially.
5 The very nonunifonn flow profile in two-phase flow in
inclined pipes is believed to be the major difficulty for
the production logging tools tested. Even though the
tools may be measuring accurately the velocity, density,
4 or water fraction at the point in the wellbore at which
they are placed, the nonunifonn flow regime in deviated
wells results in erroneous extrapolation of these
3 measurements to the entire pipe diameter. A film 3 made
during the production logging tests graphically illustrates
these flow regimes and the reason we need tools that can
c:>
Q)
o sample the entire flow-stream cross section .
tJ)
~--------f_~~~------------~x Another indication of nonunifonn velocity profiles
U,
'E was obtained with Spinner Tool B and power cen-
~
0
. tralizers. These devices were used to measure localized
>- velocities of two liquid phases in a 6-in. (l52-mm) pipe
~
u at three axial positions-at the top, in the center, and at
0
...J
the bottom of the pipe. From these measurements, radial
W
>
0 velocity profiles were drawn for various flow rates,
water fractions, and inclination angles. Fig. 3 shows the
radial velocity profiles for a kerosene rate of 25 gal/min
-1 (1.6 dm 3 Is), a water rate of 25 gal/min (1.6 dm 3 Is), and
various inclination angles. Observation of these and
other velocity profiles resulted in the following general
conclusions.
-2 Angle from 1. The velocity profile is essentially flat in vertical
Vertical (degrees) flow (bubble flow regime).
x 0 2. The velocity profile becomes progressively steeper
-3 o 5 as the pipe angle changes from vertical to 45 0. The
d 15 change in slope of the velocity profile is most noticeable
c 30 between 5 and 15 from vertical.
45 3. At constant liquid flow rate and pipe angle, veloci-
-4
60 ty profiles become steeper as the kerosene fraction is in-
creased from 0.1 to 0.5 to 0.9. Because the velocity pro-
-5~ _____________________________...J file flattens out again when the kerosene fraction reaches
Bottom Top 1.0, there must be a very rapid transition in the velocity
RADIAL POSITION IN PIPE profile at a kerosene fraction between 0.9 and 1.0.
The simplest observed fluid flow situation is single-
Fig. 3-Velocity profiles in oil/water two-phase flow. phase vertical flow; the most complicated is two-phase
inclined flow with a water fraction of less than 0.1 and
unifonn and did not correspond to flow patterns found in an inclination angle of 5 to 15 from vertical. This most
vertical or horizontal flow. complex situation represents only a small change in
In gas/liquid two-phase flow, slug flow was the water fraction and angle from the simplest situation and
predominant flow regime when the pipe deviated from is very common in both onshore and offshore operations.
vertical. This flow pattern corresponds to the "both
phases continuous" flow regime identified by Gould et
at. 2 in two-phase upward flow at 45 pipe inclination. Production Logging Tool Behavior
The gas slugs would move rapidly up the upper side of Four main types of tools from the three companies were
the pipe, and a reverse liquid flow usually was observed tested: spinner flowmeters, percent water tools, density
along the lower side of the pipe. Thus, the velocity pro- tools, and flow-concentrating flowmeters. All three
file across the pipe was very nonunifonn (Fig. 2), mak- companies provided the first three types of tools. Two
ing spinner-type velocity measurements difficult to relate types of flow-concentrating flowmeters were tested.
to actual volumetric flow rates of the two phases.
Because the gas occupied a small portion of the pipe Spinner Tools
cross section, density tools positioned in the middle or Spinner flowmeters from each of the service companies
lower part of the pipe could not be expected to provide a were tested in the experimental flow facility. Spinner A
true average flow-stream density. is a four-blade spinner 3 V2 in. (88.9 mm) in diameter.
In oil/water two-phase flow, a segregated flow regime Both Spinners Band C are helical spinners 1 liz in. (38.1
was common, with the denser phase again occupying mm) in diameter.
2434 JOURNAL OF PETROLEUM TECHNOLOGY
These flowmeters are designed to measure relative
4 FLOW RATES
fluid movements through specific zones of interest in ~
cased or open wells. Under conditions of turbulent Ul 850 BID Oil O
e- 7.9 MCF/D Gas ...J
single-phase vertical flow, the rotational velocity of the W
LA..
a..
spinner is approximately proportional to the velocity of en :J
z
fluid movement. However, multiphase flow in a 0
a..
en
t
deviated well bore produces a more complex flow w 0
situation. a::
a:: ~
+
Fig. 2 shows a typical gas/liquid flow situation. The w 0
Z ...J
gas phase passes in slugs along the top of the pipe at a Z LA..
much higher velocity than does the liquid phase that oc- a::
en
z
~
cupies most of the pipe volume. As gas slugs move -4 0
0
rapidly up the top of the pipe, the liquid moves in large
turbulent eddies at a much slower velocity. Average
fluid velocity decreases rapidly from the top to the bot- Fig. 4-Spinner tool response as a function of pipe inclina-
tom of the pipe, and, under many flow conditions, fluid tion in two-phase flow.
velocity along the bottom of the pipe is continuously
downward.
Because of the complex velocity profile, spinner
flowmeter response is very difficult to interpret in in-
clined multiphase flow situations. Depending on the
spinner positioning, the response can vary from a very
fast upward velocity along the top of the pipe, to erratic
spinner motion as the fluid movement is alternately up-
ward and downward in the center of the pipe, to a steady 2000 BID
down response of the spinner along the bottom of the
pipe. Spinner tools measure only a localized velocity at KEROSENE
one position in the wellbore. In complex flow, it is im- 7.5 Mcf/d GAS
possible to determine total fluid flow rate from a single
localized velocity measurement.
30
Fig. 4 shows the effect of pipe inclination on Spinner
C for a specific combination of fluid flow rates. Note
that the effect of pipe angle on tool response changes as
the tool is moved from the center of the pipe to the bot-
tom. As shown in the figure, when the tool is central-
ized, the response does not change greatly between ver-
tical and a deviation of about 10 from vertical. In con-
2000 BID KEROSENE
trast, when the tool is not centralized, the response 1.5 Mcf/d GAS
changes drastically at very small deviations from 30
vertical.
The difference between spinner tool response in
single-phase and gas/liquid two-phase flow in an in-
clined pipe is illustrated in Fig. 5, a strip-chart recording
of instantaneous spinner response. For single-phase
kerosene flow, the spinner yields a fairly smooth upflow
response. When gas flow is added at a rate of 1.5 Mcf/D
(42 m 3 /d), the spinner response becomes very erratic, Downflow
reflecting the rapid reversals in spin direction taking ~
place. The response also has shifted so that an integrated
signal is indicating a net downflow. An increase in gas
rate to 7.5 Mcf/D (210 m 3 / d) increases the fluctuations
in spinner response. 2000 BID KEROSEN E
Percent-Water Tools NO GAS
Tools for measuring the fraction of water in the flowing 30
stream were tested from each of the logging companies.
The tools are essentially coaxial capacitors that infer the
water fraction by measuring the fluid's dielectric con-
stant or capacitance. With Percent-Water Tool C, fluids
enter and leave the capacitance measuring chamber
through circular ports. With Percent-Water Tool B, fluid
flows through the tool via large slots. Both devices allow
the fluid being sampled to flow continuously through the Fig. 5-Strip chart record of spinner tool response in two-
tool. Percent-Water Tool A differs in that it restricts flow phase inclined flow.

OCTOBER 1982 2435


I I I I I be compared to the actual percent water present in the -
o Stratified Kerosene-Water Mix test section as ascertained from fl.p measurements.
o Oil-Water Emulsions Because the frictional pressure drop is low compared
11,200 f- - with the gravity head, a ~p measurement is a good in-
00
0 dication of the average density of fluids in the pipe.
0 0 However, analysis of the ~p measurements shows them
10,800 I- 0 - not to be accurate enough in kerosene/water two-phase
flow for use as a valid comparison. Thus, only
N 0
qualitative observations can be made about the behavior
I 0
~ 10.400 - - of these tools in the flow tests.
en 0
The main difficulty with the percent-water tools was
Z 0
0 that in kerosene/water two-phase flow, they usually were
55 10,0001- 00 - immersed in the water phase when the pipe was deviated
w c from vertical. Because of their location, they indicated
a: 0
...J only the water that is present in the pipe and thus did not
C
0 0 accurately reflect the true percentage of water in the
0 9600 I- 0 -
I- flowing stream.
0
Density Tools
9200 I- - Three density-measuring tools were tested. Density Tool
0 c 0
0
A measures the fluid head across 2 ft (0.6 m) in the
wellbore. Density Tools Band C measure the gamma ray
8800 I I I I ~ absorbance of the well bore fluid.
0 20 40 60 80 100
The gamma ray density tools have a gamma ray source
WATER (%) on one side of the sampling channel and the detector on
the other. Density measurements are made by observing
Fig. 6-Percent-water Tool A calibration.
the relative absorbance of gamma rays passing through
the wellbore fluid in the sampling chamber. The count of
residual gamma rays at the detector is inversely propor-
tional to sample density.
into the measurement chamber; two small holes provide The major problem with gamma ray density tools is
the only communication with the flow stream. Accurate that they do not measure a bulk density of the well bore
measurement depends on gravity separation of the oil fluid but rather a localized density of the fluid in the sam-
and water in the tool. Ideally, the upper portion of the ple chamber. This is similar to the localized measure-
capacitor will be immersed in oil and the lower portion in ment problem observed with the percent water tools.
water. The tool thus can sense the fluid's average dielec- Density Tool A measures the difference in pressure in
tric constant. 2 ft (0.6 m) of well bore , which is related to the bulk den-
Calibration in static fluids was the first step taken in sity of the wellbore fluids. The pressure drop is
testing these tools. In the field, the tools generally are measured by the mechanical shifting of a pair of bellows.
calibrated by obtaining one measurement for oil and one Corrections must be made for well bore deviation as well
for water, and by assuming a linear dependence of tool as frictional and kinematic pressure components.
response on water fraction between these endpoints. The short length over which the pressure drop is
Two calibration techniques were used in this study to test measured reduces the precision of Density Tool A.
the percent-water tools. In the first technique, a series of Limited precision is most critical in oil/water detection
mixtures of oil and water was formulated in 10% in- when the density difference between the two phases is
crements of water fraction. A manometer oil with a small. The 2-ft (O.6-m) length is also too short to smooth
specific gravity of 1.0 was used in the mixtures so that out the slugging effects of gas/liquid flow regimes,
they would be stable for a long time. Fig. 6 shows the although mechanical dampening is incorporated in the
calibration curve for Percent-Water Tool A with these tool. Density Tool A is not affected adversely by a
mixtures. Two linear responses were obtained-one cor- stratified flow regime, since the ~p measurement will
responding to mixtures in which oil was the external represent average density regardless of the flow pattern.
phase, and another corresponding to water-external
mixtures. Flow-Concentrating Flowmeters
The second calibration method was a linear calibration Two tools that attempt to direct the entire flow stream
in which the tools were immersed in a tank containing through a small spinner flowmeter were tested. Both
kerosene floating on water. The tools were immersed to devices provided a better measurement of total
various depths below the kerosene/water interface so that volumetric flow rate than did any conventional spinner
a fraction of the tool was exposed to kerosene while the tool.
remainder of the tool was exposed to water. The linear Flow-Concentrating Tool A is a 2Vs-in. (54-mm) tool
type of calibration for Percent-Water Tool A produced a consisting of an inflatable rubber bag and a small-
linear response through the range of 30 to 95 % water. diameter turbine inside the tool housing. In operation,
The percent-water tools were tested over a wide range the rubber bag is filled with well bore fluids by a small
of one-, two-, and three-phase flow conditions. We had pump, which forms a constriction in the wellbore. The
hoped that the percent water predicted by the tools could fluids are forced into ports below the packer, pass
2436 JOURNAL OF PETROLEUM TECHNOLOGY
through the turbine, and exit from ports above the
packer.
Flow-Concentrating Tool A was operated at flow rates
as high as 2,760 BID (425 m 3 /d) of liquid with 1.2
McfID (34 m 3 /d) of gas and 2,520 BID (401 m 3 /d) of w
00
liquid with 5.5 McfID (156 m 3 /d) of gas. At these rates, Z
pressure drops as high as 14.35 psi (98.9 kPa) were o
a..
measured across the tool. That pressure drop translates to 00
w
a force of 406 Ibf (1810 N) on the packer in the 6-in. a:
...J
(152-mm) ID test pipe. In the experiments, the tool c{
string was clamped to the flange at the top of the test sec- r-
tion to prevent movement of the tool up the pipe. In a
or-
field logging operation, sufficient weight would have to
be added to the total string to overcome this force to
make measurements at these flow rates.
Flow-Concentrating Tool A responded linearly to total
volumetric flow rate over a wide range of flow condi-
FLOW RATE (gpm)
tions. Fig. 7 shows the tool's responses to single-phase
flow of kerosene or water, two-phase flow of kerosene Fig. 7-Flow-Concentrating Tool A response in one-, two-,
and water, and two- and three-phase gas/liquid flows. and three-phase flow.
These figures all contain data taken over a range of well
deviations from 0 (vertical) to 60 from vertical. As ex-
pected, Flow-Concentrating Tool A is not affected
significantly by well angle. 125
On the basis of these results, this tool appears to be a
good device for measuring volumetric flow rate. Its
100
drawbacks are its upper range limit of about 2,000 B/D N
(328 m 3 /d), fragility of the rubber bag under field condi- I
UJ
tions, and that it can be used only between perforations. (f)
z 75
However, where it can be used, it should give much 0
Cl.
more reliable measurements of total volumetric flow (f)
UJ
rates than any spinner device. a: 50
Flow-Concentrating Tool B is based on the same prin- ....J
0 0
ciple as Flow-Concentrating Tool A-it attempts to 0
I-
direct the entire flow stream through a small spinner 25 0 Experimental Data
flowmeter. It differs in that a metal funnel instead of a - Least-Squares Fit
packer is used to force the fluids through the spinner.
0
The funnel is constructed of metal petals attached to cen- 0 125
tralizer bands. When the tool is in position for logging,
the bands are compressed, opening the funnel to the
diameter of the pipe.
Fig. a-Flow-Concentrating Tool B response in single-phase
Before our tests, Flow-Concentrating Tool B had liquid and two-phase oil/water flow.
never been tested in flowing fluids. One obvious flaw of
the tool noted in initial testing was the fragility of the
funnel petals, which originally were made of beryllium-
copper sheet metal. When this material was replaced
with stainless steel, the tool withstood much higher flow
rates before it was damaged. With the stainless-steel
petals, this tool was operated successfully at flow rates 3
as high as 2,640 BID (420 m 3 /d) without gas.
The response of Flow-Concentrating Tool B to total
flow rate was fairly linear to single-phase liquid or two-
phase kerosene/water flow (Fig. 8). When two-phase z 2
~
gas/liquid data are added, there is still a linear trend, but o Least-Squares Fit
o
the data are more scattered. Since these tests, Flow- d.. Slope = 22.2 fpm/rps
Concentrating Tool B has been field tested successfully => o Intercept = 4.4 fpm
in a well in the McArthur River field. The well had a
deviation angle of 38 from vertical.
Statistical Analysis of Tool Responses O~~ __- L______ ~ ____---'______- L____~
Stepwise, linear regression analysis was used to correlate o 25 50 75 100 125
responses from two of the tool strings to the flow KEROSENE RATE (gpm)
parameters. We attempted to correlate the density, per-
cent water, and spinner flowmeter tool responses to Fig. 9-Spinner Tool C single-phase response.

OCTOBER 1982 2437


6 functions of the oil, water, and gas flow rates and the
deviation angle. With a correlation for three tools, the
5 equations obtained can be inverted to predict the flow
rates of oil, gas, and water, given the tool responses. The
4
correlations obtained, however, are not sufficiently ac-
Cii 3 curate for use as predictive equations. The analysis in-

z
e-
~~<
;:::::
. . ",/ . . dicated that tools are required that measure averaged
properties of the flow stream. Even though no equations
3:
o ----- ..::.~ that are ready for field adaptation are developed, the
o,
.;~.., potential of this analysis method is demonstrated, and

.
a.. --' . Pipe Angle
::> 0 . ..----- / (degrees) necessary improvements in tool designs and data acquisi-
/-,' y 0 0 tion methods are pointed out.
~.. "'~~ 6 5 A detailed examination of one of the tool string's
-1
..,~.,
-'- -::::;:./ x 15
responses was undertaken to develop an understanding
-2 ~-+<~/- 30 of relationships between tool response and observed
+ 45
D 60 fluid flow rates. From these relationships, correlations
25 50 75 100 125 were developed to predict fluid flow rates, given the
KEROSENE RATE (gpm) tools' responses. This approach also was applied to the
Fig. 10-Spinner Tool C two-phase response.
second tool string. The statistical package used was the
Biomedical Computer Program (BMDP) available from
the U. of California. 4
Data from Tool String C were analyzed first and most
5
extensively. Several data crossplots were generated to
Slope = 24.6 fpm/rps develop relationships between flow parameters and tool
4 responses. With an understanding of these relationships,
a regression analysis of the data was performed to
3 develop equations of tool responses as functions of flow
parameters. These fitted equations then were used to
calculate predicted flow rates, given the tool responses;
these values were compared with the observed flow rate.
Fluid velocities were measured with Spinner Tool C.
The observed tool response was correlated to observed
fluid flow rates. Spinner Tool C was used to observe
velocities both at the center and at the bottom of the pipe.
-1 Because spinner tool measurements in the center of the
pipe were found more consistent and more representative
of the true flow situation than measurements taken at the
bottom, only centralized spinner data were used for
statistical analysis of tool response.
50 75 Hio 125
Fig. 9 shows Spinner Tool C response for flow of a
WATER-KEROSENE RATE (gpm)
single fluid phase at pipe angles from 0 to 60 from ver-
Fig. 11-Spinner Tool C three-phase response. tical. This plot shows a straight-line relationship between
flow rate and spinner response, independent of pipe
angle. The line intersects the flow rate axis at a small
Pipe Angle
5 positive value corresponding to the flow rate required to
(degrees) 50 gpm Kerosene
0 0
overcome bearing friction.
~ 5 Flow becomes much more complex with the introduc-
4 x 15 tion of a second phase, as shown in Fig. 10 for a
30 kerosene/gas system. With the presence of two phases of
+ 45 differing densities, the angle of inclination has a strong
Cii 3 o 60
effect on spinner response (Fig. 4). However, at a given
~
z angle of inclination and a fixed gas rate, tool response is
#
3: 2 linear with respect to kerosene flow rate. This relation-
0
0 ship is shown in Fig. 10 for six angles of inclination and
!l. a constant gas rate of 7.9 Mcf/D (224 m 3 / d).
::>
~. Spinner Tool C response is recorded in up rotations
~----~---------- .... minus down rotations. Other functions of flowmeter
Or---~=-=------~-'~--=c;~~~--~ response, such as the sum of up and down rotations, are
-==.:to
not linear with respect to liquid flow rate. This distinc-
-1~_~ _ _~_ _~_~___~____~__~ tion is most significant at low liquid and high gas rates
o 2 3 4 5 6 7 when reverse rotations are substantial.
GAS RATE (cfm) Fig. 11 shows Spinner Tool C responses for three-
phase flow with a water fraction of 50% of total liquid
Fig. 12-Spinner Tool C dependence on gas flow rates. rate. The flowmeter response was linear with respect to
2438 JOURNAL OF PETROLEUM TECHNOLOGY
total liquid rate for a constant gas rate and angle of in- OK CALCULATED (gpm)
clination. Apparently, in three-phase flow, the two liq-
uid phases emulsified and behaved as a single phase.
This is also true for water fractions of 10% and 90%.
The effect of gas rate on flowmeter response also was
studied. Fig. 12 shows the change in flowmeter response
with gas rate at varying angles and a constant liquid
(kerosene) rate. From the appearance of this plot and o
;,.;:
similar plots of other rates, the following function was
001
chosen to fit the effect of gas rate on flowmeter response. OlO
C/l
m
:D
<
m
0 ....
COOl
If only the positive square root is used, this equation "0

defines a half ellipse that inverts from the top half of the 2.
ellipse to the bottom half as the angle increases through
15 .
With this information, the flowmeter response was fit-
ted to the three fluid flow rates and to the angle of in-
clination by use of the stepwise regression analysis pro-
gram. The resulting least squares equation is
Fig. 13-Comparison of predicted and actual kerosene flow
Rv = -0.528+0.339qk +0.0308qw rates for Tool String C correlation.

+0.128qg -1.90 sin(O-15)qg 1,\ ........ (2)


Similarly, the response of the percent-water tool was
r 2 =0.82, fitted to the observed data to give the equation

where R v is the flowmeter response in revolutions per qk


second, qk and qw are the kerosene and water flow rates Rwc = 11,300-64.4qw +9,100---
in gallons per minute, * q g is the gas flow rate in cubic qk+qw
feet per second, ** and 0 is the angle in degrees from ver-
tical. The r2 value is the square of the adjusted coeffi- _ 1 880 q k +q w ............ (4)
, qk+qw+7 .48qg'
cient of correlation that indicates how well the equation
fits the data. A rough way of interpreting r2 =0.82 is to
say that 82 % of the variance in the data is predicted by r2 =0.79,
the correlation.
Density Tool C and Percent-Water Tool C were where Rwc is the response of the percent-water tool.
analyzed similarly. The density tool response was fitted Gas flow rate and angle of inclination have an in-
to the fluid flow rates by use of regression analysis. Bulk significant effect on percent water tool response because
fluid density is affected strongly by the fraction of liquid the tool detects only a localized fluid property and does
flowing in the pipe as well as the fraction of water flow- not detect the gas at the top of the pipe.
ing in the liquid phase. Adding these two variables to the The three equations of tool responses as functions of
list developed for the flowmeter, we developed the fluid flow rates can be transformed to solve for the three
following equation. fluid flow rates as functions of tool responses. Ideally,
these three equations could be used with production log-
ging results to define fluid flow rate at specific points in
the wellbore.
Starting with Eqs. 2, 3, and 4, it is not possible to
solve for the three fluid flow rates analytically.
-63.0 qk+qw Therefore, the equations were solved numerically for
qk +qw +7.48qg each set of logging tool responses to obtain values for
fluid flow rates. The calculated values were plotted
-38.45 sin(O-15)qg v>, ................. (3) against observed fluid flow rates that correspond to each
set of tool responses. Fig. 13 is one such plot, comparing
the kerosene flow rate calculated from the statistical cor-
r2 =0.92, relation with the flow rate measured experimentally. Ob-
viously, the correlation is not sufficiently accurate to use
where Rp is the response of the gamma ray density tool. in interpreting production logs.
The larger r2 value indicates that this equation fits the We conducted a similar regression analysis of another
raw data fairly well. production logging tool string and transposed the
resulting equations to provide analytical expressions of
'1 gal/min = 6.3 dm 3 Is. fluid flow rates as functions of tool response. Analysis of
"1 cu ftlsec=2.8 dm 3,s. these data produced the following regression equations.
OCTOBER 1982 2439
Rw = -20.45+0.266qk +0.302qw downflow along the lower side of the pipe, which results
in misleading tool responses.
+17.4 cosO, .......................... (5) 3. The flow-concentrating-type flow meters are the
most reliable tools for measuring flow rates in deviated
r2 =0.76. wells.
4. Capacitance tool responses are not linear functions
qk
Rp = 1.50-0.812 cosO-0.116--- of water fraction in calibrations in static fluids. The im-
qk+qw precise calibration makes interpretation of these tools'
responses in flowing fluids difficult.
+0.340 qk +qw , ............... (6) 5. Regression equations of production logging tool
qk +qw +7.48qg responses did not predict the observed fluid flow rates
accurately. Logging tools must be redesigned to measure
r2=0.94. bulk flow properties before production logs can be inter-
preted reliably in the field.
qk
Rwc =8,420+ 746 cosO+29.1qw +2,910--- Nomenclature
qk+qw
CI a constant
=
-523 qk +qw , ................ (7) qg gas volumetric flow rate, cu ft/D
=
q k + q w + 7 .48q g qk = kerosene flow rate, gal/min
qL = liquid flow rate, gal/min
qw = water flow rate, gal/min
R fc= flow concentrating tool output
Rv = spinner tool output
The very low r2 value for the percent-water tool in- R we = percent water tool output
dicates that there is essentially no correlation between R p = density tool output
tool response and observed fluid flow rates. 1p = pressure difference
In the analyses of both tool strings, the percent-water
type of tool was found to be the most difficult to cor-
o = inclination angle, measured from vertical
relate with observed flow parameters. Improvements in
tool design will be needed before the percent water tool Acknowledgments
becomes a reliable device for use in the field.
We thank the management of Marathon Oil Co. for per-
The results of regression analysis can be improved mission to publish this paper. We also express our ap-
dramatically by a change in tool design. Flow-
preciation to the many individuals at the three service
Concentrating Tools A and B are much more correlatable
companies involved whose assistance and cooperation
with observed fluid flow rates than the previously
contributed greatly to this work.
discussed flowmeters. Analysis of the data for Flow-
Concentrating Tool A resulted in the regression equation
References
R fc =0.066+0.097qg +0.021qL, ............. (8) I. Govier, G.W. and Aziz, K.: The Flow of Complex Mixtures in
Pipes, Van Nostrand Reinhold Co., New York City (1972).
2. Gould, T.L., Tek, M.R., and Katz, D.L.: "Two-Phase Flow
Through Vertical, Inclined, or Curved Pipe," J. Pet. Tech. (Aug.
r 2 =0.97, 1974) 915-26; Trans., AIME, 257.
3. Hill, A.D., Johnson, I.D., and Oolman, T.: "Effects of Two-
Phase Flow Regimes in Deviated Wells on Production Logging
where R fc is the voltage response from the tool, and Tool Response," paper SPE 9048 presented at the 1980 SPE
qL=qk+qw' This is a much simpler and better-fitting Rocky Mountain Regional Meeting, May 14-16.
equation than those for the spinner flowmeters (Eqs. 2 4. BMDP Biomedical Computer Programs, Dept. of Biomathemat-
and 5). The entire fluid stream flows through Flow- ics, U. of California, Los Angeles (1975).
Concentrating Tool A, resulting in a measurement of
bulk fluid velocity instead of a localized velocity.
SI Metric Conversion Factors
Conclusions bbl x 1.589 873 E-Ol m3
1. Flow patterns in multi phase flow in inclined pipes cu ft x 2.831 685 E-02 m3
dramatically affect production logging tool responses ft x 3.048* E-Ol m
and prevent the devices from accurately measuring gal x 3.785 412 E+OO dm 3
volumetric flow rates, average densities, or holdups. Conversion factor is exact. JPT
2. Spinner-type flow meters work wel~ in single-phase Original manuscript received in Society of Petroleum Engineers office July 20. 1981.
flow at any well angle. In two- or three-phase flow, as Paper accepted for publication June 8. 1982. Revised manuscript received Aug. 2,
1982. Paper (SPE 10208) first presented at the 1981 SPE Annual Technical Con-
little as 2 of deviation from vertical can cause
0
ference and Exhibition held in San Antonio Oct. 5-7.

2440 JOURNAL OF PETROLEUM TECHNOLOGY

Anda mungkin juga menyukai