7th International Conference on Multiphase Flow
ICMF 2010, Tampa, FL USA, May 30June 4, 2010
Flow Rate Measurement of Slug Flow Based on DualPlane CrossSection Resistance
Information Measuring System
Fusheng Zhang and Feng Dong
Tianjin Key Laboratory of Process Measurement and Control, School of Electrical Engineering and Automation,
Tianjin University, Tianjin 300072, China
Email: fdonwi; ilju i cdu en
Keywords: Gasliquid twophase flow, slug flow, crosssection resistance information measuring system, flow rate
Abstract
Gas/liquid two phase flow especially the slug flow receives great attention in the field of multiphase flow research for its
conmen appearance in many industry processes. Researchers worldwide apply themselves to the work of slug flow generation,
transmission and parameter measurement. However the present measurement methods can not fully fill the industrial
requirement. In this paper, an electrodeelectrode crosscorrelation method is put forward to make the mixed velocity
measurement of slug flow. The experiment is carried out in a 50mm diameter horizontal pipe, the dualplane system used is
16electrode structured with a speed of 450 frames/s for each measuring plane. Excitation manner has been sorted into three
kinds: water excitation (WE), gas excitation (GE) and Interface excitation (IE) based on the regular phase distribution of slug
flow and the position of exciting electrodepair. Different from the traditional data preprocess, single electrode analysis method
proposed in this work gives more sensitive and correct results. The experiment error of electrodeelectrode crosscorrelation
method is within 10% and acceptable.
Introduction
Slug flow is an important flow regime of gasliquid
twophase flow. The parameter measurement of slug flow
receives great attention in both scientific and engineering
field for its conmen appearance in many industry processes
such as petroleum industry, chemical industry, metallurgical
industry, nuclear industry and so on (Hewitt, 1978).
Researchers worldwide applied themselves to the slug flow
generation, recognition, transmission and parameter
measurement. The accuracy of the slug flow flowrate
measurement method exists nowadays can not fully fill the
requirement of industrial processes.
The measurement methods exist for flowrate measurement
include Coriolis meter (Shanmugavalli and Umapathy et al.,
2010), differential pressure (DP) meter with twophase flow
models, such as orifice plate DP meter (Murdock, 1962),
Venturi meter (Lin, 1982) and VCone meter (Steven, 2002),
volumetric meter (Thorn and Johansen ea al., 1997),
correlation technology (Xu, 1988), ultrasonic technique and
so on. Crosscorrelation technique as an important method
of flowrate measurement receives great attention in the field
of multiphase flow (Lucas and Jin 2001, Dong and Xu ea
al., 2005). The mathematical fundamental of
crosscorrelation technique is stochastic process. By
analysing the stochastic noise signal of two homogeneous
sensors installed with the same axis, crosscorrelation
technique turns flowrate measurement into the transfer time
measurement between the two sensors. The homogenous
sensor could be optical, acoustic, electrical sensor .etc.
Crosscorrelation method based on electrical resistance
tomography (ERT) takes electrical sensor, electrode arrays,
as its correlation sensor. The system usually has two
electrodeplanes respectively as upstream and downstream
sensor. The ERT based crosscorrelation mainly contains
two generic data dealing methods. The first is pixelpixel
correlation after cross images had been reconstructed (Deng
and Peng et al., 2004). The second is featurefeature
correlation, in which the feature is extracted out to represent
the phase distribution of the crosssection (Dong and Xu ea
al., 2005). The pixel correlation has the advantage of
detailed analysis to the crosssectional phase distribution
but with the limitation of image reconstruction accuracy
which remains a difficult issue in the corresponding
research field. The feature correlation method calculate the
voltage date directly with no need of image reconstruction,
however, it take the whole electrode plane as a correlation
sensor without any detailed information in the crosssection
analyzing ability according to the pixel correlation method.
A crosscorrelation method based on crosssection
resistance information is put forward in this paper to make
the mixed velocity measurement of slug flow. Different
from above crosscorrelation method, the correlation
calculation is directly carried out between the voltage data
of corresponding electrodes of the two electrode planes. It is
designed to have both merits of the non image
reconstruction and detailed information analyzing ability. In
this method, the exciting electrodes are sorted into three
7th International Conference on Multiphase Flow
ICMF 2010, Tampa, FL USA, May 30June 4, 2010
kinds based on their crosssectional position which include
electrode in gas, electrode in water and electrode at the
interface of them. Measured data are further sorted into
three kinds for each exciting style according to the
measured electrodes' crosssectional position with the same
rule of exciting electrode classification. Correlation velocity
behavior of the above 9 kinds of electrode voltage data is
disused. With consideration to the velocity resolution of
correlation velocity, only 8 electrode pairs are needed to
participate in excitation process instead of 16 pairs in
conventional method. This character has improved velocity
resolution by 2 times in the same hardware condition.
Correlation velocity is related with mixed velocity, and gas
quality for gas/liquid twophase flow. Because the gas
quality is under 0.05 in the experiment, only small impact
on correlation velocity occurs, so relation between gas
quality and correlation velocity is out consideration. Only
the relationship of mixed velocity and correlation velocity is
discussed in this paper. The experiment result shows that
this crosscorrelation method is efficient for mixed velocity
measurement of slug flow.
Nomenclature
coefficient matrix
distance between the planes (m)
pressure (Pa)
volt (v)
final correlation velocity (m/s)
velocity resolution (/n)
Greek letters
6 system acquisition time for each frame (frame/s)
CrossSection Resistance Information Measuring
System
The crosssection resistance information measuring system
uses the same hardware platform with ERT technique and its
function is not limited to the image reconstruction. The
system could realize the phase distribution analysis by
image reconstruction or make the flow regime recognition,
parameter calculation right through the direct analysis of the
electrode voltage data either. The physics fundamental of
resistancial crosssection information measuring system is
that different medium has different conductivity. The change
of medium distribution in the crosssection leads to the
change of crosssectional conductivity distribution which
can be measured by the system (Tan and Dong et al., 2007).
Adjacent excitation manner is usually ultilized in
resistancial crosssection information measuring system
shown in figure 1. Different from the working process of
exgeneration parallel system (Tan and Dong et al., 2007),
the voltage data of electrodes are directly and
simultaneously measured in one exciting process, while
voltages between adjacent electordes are measured in the
exgeneration parallel system. Exciting current is then
injected into the next pair of adjacent electrodes until all the
adjacent electrodes have been excited. Except the voltages
of exciting electrodes, a frame of crosssection information
can be represented by 16 x14 voltage data.
Exciting Switching Voltage
*4 r
'Ibi'
Exciting current
Figure 1: System working principle and electrode
structure
A dualplane crosssection resistance information measuring
system (Zhang and Dong et al., 2010) with 16electrode
architecture is used in this work. 16 electrodes are uniformly
distributed in each measuring plane and the distance
between two planes is 10mm shown in figure 1. PXI (PCI
extensions for Instrumentation) modular instruments and
FPGA technique have been adopted in the system to
improve the realtime parameter. The data acquisition speed
is up to 450 frames/s for each electrode plane.
Signal Processes
The data structure of crosssection resistance information
measuring system for each frame is 16x16 including the
voltage data of exciting electrode pair. The data dimension
is large and hard to be calculated or be related with flow
parameter. In traditional system, the voltage data of exciting
electrode pair are not acquired because the voltage value is
much large than the data of other electrodes and the
hardware acquisition ability is limited. To reduce the data
dimension the following data reprocess is carried out. 14
measured voltage data except the exciting electrodes data
are made adjacent subtraction in one pair exciting process.
Thus a data structure of 16x 13 is formed for a frame. F, is
extracted out to represent the feature of one excitation.
=13A )
Where v is the j th voltage data at the i th excitation,
V is the V value when the pipe is full of water.
Figure 2 shows a F, series of 500 frames in length. Slug
flow recognition could be efficiently carried out based on
F, series (Zhang and Dong et al., 2008). The F, section
with high value represents the gas slug of slug flow and F,
section with a small value represents the liquid slug.
200
a2 m
200
20
0" ^
1.00 H ^^ ^
F, Number
Figure 2: F, series of slug flow
7th International Conference on Multiphase Flow
ICMF 2010, Tampa, FL USA, May 30June 4, 2010
With fast technology development, acquisition device with
high faster speed and wider range such as PXI platform
appears. The voltage data of exciting electrode pair is also
fetched in the system used in this paper. The 16x16 data for
each frame is sorted into two parts: 16x2 exciting part and
16x14 nonexciting part. The former could be used for slug
flow recognition and the latter part is directly used in the
crosscorrelation calculation.
Data Classification
Liquid and gas phases are regularly distributed in slug flow
of horizontal pipeline. The body of a gas slug can be
considered as stratified flow, and the head and tail of gas
slug as the time integration of stratified flow. Since slug
flow has above regular phase distribution, the electrode pair
in excitation is cut into three manners based on their
crosssectional position, which include under water
excitation (WE), exposed in gas excitation (GE) and
Interface excitation (IE). Measured electrodes are sorted
into three kinds in a further step for each excitation based on
the same position rule, shown in figure 3. In this way,
measured voltage data in each frame has been divided into
R, (i = 1 ..9 ) with different position in the crosssection.
Gas
Excitation
9 Interface
Excitation
2 Water / 7 Water
3 6 Excitation
4 5
Figure 3: Electrode classification in excitation and
measurement
R, (i = 1..9 ) series of slug flow for 1000 frames in length
is shown in figure 4. R, ( i=1.9 ) have different
responses to the gas slug and liquid slug and have good
differentiating ability from each other.
20 R1 measured electrode in water
R2 measured electrode in gas
R3 measured electrode at interface
15
> ')
Frame Number /N
(a) Measured voltage in water excitation
5 20
S15
0
4 measured electrode in water
5 measured electrode in gas
R measured electrode at interface
Frame Number /N
(b) Measured voltage in gas excitation
 R7 measured electrode in water
 R8 measured electrode in gas
 Rg measured electrode at interface
11 III I *
:IY TP LJ
0 250 500 750 1000
Frame Number IN
(C) Measured voltage in interface excitation
Figure 4: V, series in WE, GE and IE
CrossCorrelation Technique Based
Crosssection Resistance Information
on the
In the traditional crosscorrelation calculation based on
crosssection resistance information, the whole electrode
plane has been considered as a single correlation sensor. R,
series from upstream and downstream plane are respectively
defined as x(t) and y(t). Correlation function R~ (r) is
defined as follows in order to analyze the similarity degree
of x(t) and y(t).
T
R, ()= lim x(t)y(t+ r)dt (2)
T4 *
0
The peak value time of correlation function is called
transmit time (To). Since the distance between the two
electrodeplane is already known, the correlation velocity
calculation has been transferred into transmit time To
calculation. Correlation velocity equation is as follows.
v= L (3)
/1o
7th International Conference on Multiphase Flow
ICMF 2010, Tampa, FL USA, May 30June 4, 2010
Where v is correlation velocity, L =O.lm is the distance
between the two electrode planes.
In fact, the electrode in each plane has it own sensitivity
according to their position and can be considered as single
sensor in the crosscorrelation process. The conventional
equilibrating method has removed the sensitivity of single
electrode. A crosscorrelation measured based on the
measured voltage data of single electrode, such as alo with
blo shown in figure 1, is put forward in this paper. vi, of
upstream plane and V2, of downstream plane are defined
as x(t) and y(t) then used in (2) and (3) to make
primary correlation velocity R,.
Only 8 electrode pairs participate in excitation instead of all
the 16 electrode pairs in this electrodeelectrode
crosscorrelation method. WE and GE parts are used in the
head and tail of gas slug. Bottom electrodes no.3 to no.6 are
used as exciting electrodes in WE condition which means
electrode pairs of 34, 45, 56 are successively as electrode
pair with excitation, shown in figure 3. Upper electrodes
no.11 to no.14 are used as exciting electrodes in GE
condition. The reason is the liquid level of crosssection in
slug flow dynamic experiment is always above 1/3 of inner
diameter of the pipe, electrode no.3 to no.6 are always under
water and no.11 to no.14 are mostly likely to be exposed in
gas of the slug body. Electrode no.l, no.16, no.8 and no.9
are most likely appeared in the gasliquid boundary area in
the gas slug body, so no.1 and no.16, no.8 and no.9
excitation are used as IE. Detail velocity information could
be released by analyzing the primary correlation velocity
performance of V,. Figure 5 show electrode correlation
velocity of GE, WE and IE. The red line is inlet mixed
velocity (U, = 1.92) of gasliquid twophase flow in the
experiment.
 16.&1 electrodes excitation
. S&9 electrodes excitation
6 i
R*'
4 R'
2 I
0
1 2 3
R. R
4 5 6 7 8 9 10 11 12 13 14 15 16
Elect ode Sequence Number
(c) Correlation velocity of exciting electrode in interface
area (IE)
Figure 5: Primary correlation velocity of GE, WE, IE
As shown in figure 5, the GE condition gives more stable
and accuracy measurement for mixed velocity and measured
velocity of WE condition is larger than GE. This is because
water phase occupies the bottom of the pipe all the time,
small gas bubble may cause the increase of correlation
velocity. While exciting electrode in boundary area offers a
fluctuated and large velocity measurement, aroused by
relative movement and variable interfaces between the two
phases. Measured electrodes in gas area (R3, R6 and
R, ) perform better than the electrode in other areas.
The mean value of correlation velocity in R, is extracted
out as feature velocity V1 in the corresponding area. A
coefficient matrix C could be obtained by fitting V1 with
the inlet mixed velocity of sample data. In the experiment,
15 sets of slug flow experiment data are used as sample data
to get the coefficient C The final correlation velocity is
then calculated by using (4):
Vc = CxV
3&4 electrodes excitation
 4&5 electrodes excitation
A 5&6 electrodes excitation
Where Vc is the final correlation velocity, Vfi is mean
S23
> 2 I
1.5
1 2 3
* I
H, I l,
S value of measured R,.
R.
* *
4 5 6 7 8 9 10 11 12 13 14 15
Elec rode Sequence Number
I16
(a) Correlation velocity of exciting electrode under water
(WE)
3. 5 11&12 electrodes excitation
 12&13 electrodes excitation
3 A 13&14 electrodes excitation
I 2 I I R
1. I I I I
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16
Electrode Sequence Number
(b) Correlation velocity of exciting electrode in gas (GE)
Velocity Resolution of ElectrodeElectrode Cross
Correlation
The velocity resolution of cross correlation technique based
on crosssection information is related with velocity which
is going to be measured, distance between cross correlation
sensors and system data acquisition ability. The velocity
resolution can be calculated by the following equation.
Av = v28/2L
Where Av is velocity resolution with corresponding
measuring velocity v, 6 is data acquisition time for each
frame by dual plane crosssection information system.
Based on the classification manner mentioned above, only 8
electrode pairs participate in excitation process. This
working principle effectively reduces the dimension of
every frame data. The PXI and FPGA based system adopted
in this paper can efficiently realize this working principle
and be flexibly switched to the normal working principle
and back. The data decrease would in turn reduce the impact
R
R.
"' I
of data stream to the ROM of the system, which will
prolong the successive working time.
As mentioned in the former section, a data array of 8x16
instead of 16x16 is needed to be collected for each data
frame, which means the system data acquisition is improved
by 2 times and as a consequence the cross correlation
velocity resolving ability improved by 2 times when L and
v are supposed with the same value. In this work, system
data acquisition time is 0.0022s, velocity range is
0.922.85m/s, so the velocity resolution of maximum and
minimum velocity are respectively 0.045 m/s and
1 iiim41 and the accuracy are respectively 1.6% and
+0.5%.
Experimental Facility
The experiment is carried out in Tianjin University
Multiphase Flow Laboratory. Figure 6 shows the sketch of
the experiment facility and pipelines. The pipeline made of
stainless steel which is about 20m in length with an inner
diameter of 50mm. An organic glass window is located at
the backed of the pipeline in order to observe the flow
regimes.
Water Pump
Figure 6: Sketch of experiment facility and pipeline
The flowrate of gas and water phase are measured by
standard meter after they are pumped from the tanks. The
precision of the standard meter used in the experiment is
0.5%. The two phases flow into a horizontal pipeline
through a mixing ejector to make them initially fully mixed.
Crosssection resistance information measuring system is
installed at the downstream end of the pipe where the flow
regime is well developed and flow state is more stable. The
flowrate of gas and water are controlled by adjusting of the
valves with computer. Flowrate adjustment range of water
and gas are respectively 0.0912 m3/h and 0.186 m3/h. The
inside temperature is about 220C. Fluid flows into a
coverless tank after measurement where gas goes back to
the air and water stays for recycling. A highspeed camera is
used during the experiment to record the flow state of slug
flow from side view through the organic glass window.
Bubble flow, plug flow, slug flow, stratified flow and
annular flow are common flow regimes generated by this
experiment facility. Slug flow is the measuring objects in
this work.
Results and Discussion
In the experiment, fluid velocity that can be measured is
limited by data acquisition speed of crosssection
information system. 20 sets of slug flow experiment data
7th International Conference on Multiphase Flow
ICMF 2010, Tampa, FL USA, May 30June 4, 2010
with a mixed velocity range of 0.92 2.85 m/s are used to
give the experiment result of electrodeelectrode
crosscorrelation. 15 sets data with the same velocity range
are used to fit the coefficient matrixC. Figure 7 shows the
experiment results of the electrodeelectrode correlation
method with a comparison with result of traditional
correlation method.
electrodeelectrode correlation
traditional correlation
1.0 1:5 2.0 2.5 3:0
Mixed Velocity of Slug Flow /mlh
Figure 7: Experiment result of mixed
measuement
velocity
The experiment shows that relative error is within 5% when
mixed velocity is under 2m/s. As the mixed velocity grows,
the system differentiating capability drops, which increases
the experiment error. The experiment error is within 10% on
the whole, while the relative error of traditional correlation
is larger than 5% and the velocity measured is always
smaller than the true value. The experiment result shows
that it is efficient to utilize this electrodeelectrode
crosscorrelation method to measure the mixed velocity of
gasliquid slug flow, the experiment error is smaller than
traditional crosscorrelation method on the whole.
Conclusions
Electrodeelectrode crosscorrelation method is put forward
in this paper. The electrode voltage cross correlation
velocity behavior has been analyzed. As for slug flow, the
correlation velocity measured in GE condition is more
accurate of the mixed velocity measurement than the WE
and IE condition. Measured data of electrodes in gas area
perform better than the electrode in other areas. This is
because that gas slug occupies in the upper pipe, while
water phase continuously occupies at the bottom, so GE
area electrodes are more sensitive to the change of phase
distribution.
The velocity resolution of this electrodeelectrode cross
correlation method reaches +0.045 m/s and 0.0046 m/s
respectively under the maximum and minimum measuring
velocity. The experiment result shows that relative
experiment error is within 10% on the whole and the
accuracy decreases with the mixed velocity rising. In further
experiment and research, modify of velocity impact would
be taken into consideration. If a suitable fusion method is
implicated to deal with the primary correlation velocity, the
experiment would be better.
Acknowledgements
The author appreciates the support from National Natural
Science Foundation of China (No. 50776063) and Natural
Science Foundation of Tianjin (No. 08JCZDJC17700).
References
Beck, M. S. and Plaskowski, A. Cross Correlation
Flowmeters: Their Design and Application. Bristol England:
IOP Publiishing Ltd, pp. 240 (1987)
Deng, X. Peng, L. H., Yao, D. Y, Zhang, B. F., Velocity
Distribution Measurement Using PixelPixel Cross
Correlation of Electrical Tomography, Chinese Journal of
Electronics. Vol. 13(3): 548551 (21114)
Dong, F. Xu, Y B. Xu, L. J. Hua, L. and Qiao, X. T
Application of dualplane ERT system and crosscorrelation
technique to measure gasliquid flows in vertical upward
pipe. Flow Measurement and Instrumentation,
Vol. 16:.191197 (2005)
Hewitt G F., Measurement of Two Phase Flow Parameters,
London: Academic Press, 1978
Lin, Z. H. Twophase flow measurements with sharpedged
orifices. International Journal of Multiphase Flow, Vol. 8,
No. 6, 683693 (1982).
Lucas, G. P and Jin, N. D. A new kinematic wave model for
interpreting cross correlation velocity measurements in
vertically upward, bubbly oilinwater flows. Measurement
and Science Technology, Vol. 12:15381545 (2001)
Murdock, J. W. Twophase flow measurement with orifices.
Journal Basic Engineering, Vol. 84(4):419433 (1962)
Xu, L. A. Crosscorrelation Flow Measurement Technique,
Tianjin: Tianjin University Publishing House, 1988
Shanmugavalli, M. Umapathy, M. Uma, G. Smart coriolis
mass flowmeter, Measurement, Vol. 43(4):549555 (2010)
Steven, R. N. Wet gas metering with a horizontally mounted
Venturi meter. Flow Measurement and Instrumentation, Vol.
12, No. 56, 361372 (2002).
Tan, C. Dong, F. Wu, M. M. Identification of gas/liquid
twophase flow regime through ERTbased measurement
and feature extraction. Flow Measurement and
Instrumentation, Vol. 18(5):255261 (2007)
Tan, C. Dong, F Wang, B. B. Multiplane electrical
resistance tomography system based on parallel data
acquisition strategy, Chinese Control Conference.
Zhangjiajie, Hunan, China, 391395 (2007)
Thorn, R. Johansen, G A. Hammer, E. A. Recent
Developments in ThreePhase Flow Measurement. Journal
of Measurement Science and Technology, Vol. 8(7):691701
(1997)
Xu, L. A. Correlation Flowrate Measuring Technique.
Tianjin: Tianjin University Press (1988)
7th International Conference on Multiphase Flow
ICMF 2010, Tampa, FL USA, May 30June 4, 2010
Zhang, F. S. Dong, F. Tan, C. Liu, D. Regime Recognition
of TwoPhase Pipe Flow Based on DS Evidence Theory,
IEEE International Conference on Machine Learning and
Cybernetics, Kunming, China, 1316 July, Vol. 1:100105
(2008)
Zhang, F. S. Dong, F Song, W. Xu, C. Design of cross
section resistance information measurement system based
on PXI. Journal of Engineering and Thermophysics, Vol.31,
in press (2010)
