You are on page 1of 9

Experimental Thermal and Fluid Science 79 (2016) 245253

Contents lists available at ScienceDirect

Experimental Thermal and Fluid Science


journal homepage: www.elsevier.com/locate/etfs

Error analysis of gas and liquid flow rates metering method based on
differential pressure in wet gas
Xuebo Zheng a, Denghui He a, Zhigang Yu b, Bofeng Bai a,
a
b

State Key Laboratory of Multiphase Flow in Power Engineering, Xian Jiaotong University, Xian 710049, China
Oil and Gas Technology Research Institute, PetroChina Changqing Oilfield Branch Company, Xian 710018, China

a r t i c l e

i n f o

Article history:
Received 27 April 2016
Received in revised form 11 July 2016
Accepted 12 July 2016
Available online 14 July 2016
Keywords:
Gas and liquid flow rates
Over-reading correlation
Error propagation rules
Differential pressure fluctuation

a b s t r a c t
Online measurement of gas and liquid flow rates in wet gas is of great significance in industry. When the
differential pressure meter is used to measure the gas and liquid flow rates, an over-reading correlation is
needed to correct the liquid-induced overestimation of the gas flow rate and an auxiliary correlation is
needed to obtain the liquid fraction or flow rate. With the two correlations incorporated, the gas and liquid flow rates can be calculated. In the present study, error analysis of this metering method is conducted.
The results demonstrate that metering methods based on differential pressure exhibit similar error pattern, i.e., the prediction error of gas phase is small while that of the liquid phase is large. This phenomenon can be interpreted by error propagation and the underlying physics is that the two-phase
differential pressure is mainly dependent on the gas phase but insensitive to the liquid phase. It is found
that both the mean value and the fluctuation of the differential pressure signals are able to reflect flow
rate changes of wet gas flow. However, metering methods based on the differential pressure fluctuation
present larger prediction errors than that based on the differential pressure mean value. The reason is
that the differential pressure fluctuation has much poorer repeatability compared with the differential
pressure mean value.
2016 Elsevier Inc. All rights reserved.

1. Introduction
Wet gas is a subset of gas-liquid two-phase flow, which widely
exists in industrial processes, such as oil and gas industry, nuclear
industry, and chemistry industry. According to the technical report
released by ASME, wet gas refers to gas-liquid two-phase flow with
the Lockhart-Martinelli parameter less than or equal to 0.3 [1].
Accurate online measurement of wet gas flow rate is important
for engineering and science.
As the most robust and repeatable type of flow meters, differential pressure meters have been widely applied and researched in
wet gas flows [2]. However, a problem with differential pressure
meters is that the presence of liquid in the gas flow results in a
phenomenon termed over-reading [1], which is a positive error
of gas flow rate prediction. In order to acquire the actual gas flow
rate, one common method is to use an over-reading correlation to
correct the liquid-induced error. This approach is feasible on condition that the liquid flow rate or some form of liquid fraction
information (e.g. the gas mass fraction) is supplied. In general,
the liquid fraction is obtained via an auxiliary correlation, which
Corresponding author.
E-mail address: bfbai@mail.xjtu.edu.cn (B. Bai).
http://dx.doi.org/10.1016/j.expthermflusci.2016.07.017
0894-1777/ 2016 Elsevier Inc. All rights reserved.

relates the flow rate information of wet gas with a characteristic


parameter of the flow. Thus, with the over-reading correlation
and the auxiliary correlation incorporated, the gas and liquid flow
rates can be predicted simultaneously.
Over the past few decades, many typical over-reading correlations with regard to different throttle devices have been developed,
such as Murdock correlation [3], Bizon correlation [4], Chisholm
correlation [5], de Leeuw correlation [6], Lin correlation [7] and
others, and an enormous amount of researches have been
conducted worldwide to improve the performance of these typical
correlations [813]. As to the auxiliary correlation, the key is to
extract characteristic parameters of the flow.
The across throttle device pressure drop, usually termed differential pressure, is a widely used characteristic parameter. Throttle
devices with specific geometries produce distinct differential
pressure and hence generate distinct over-reading correlations.
Therefore, by installing two different throttle devices in series
and solving the two combined over-reading correlations, the auxiliary correlation can be obtained. For example, in the work done
by Zhang [14], gas and liquid flow rates measurement was
achieved in this way. Sometimes, however, distinction between
the two over-reading correlations is very small, making the two
correlations almost identical. As a consequence, there may be no

246

X. Zheng et al. / Experimental Thermal and Fluid Science 79 (2016) 245253

Nomenclature
English symbols
A
area of the meter inlet (m2)
At
cross-sectional area at the throat (m2)
B
dimensionless parameter in Eq. (13) ()
C
error transfer coefficient ()
Cd
discharge coefficient ()
dn
coefficient in Eq. (31) ()
1
E
velocity of approach, E p
()
4

Greek symbols
q
b
equivalent diameter ratio, b AAt ()
d
deviation
dr
relative deviation (%)
DP
differential pressure (Pa)
e
expansibility coefficient ()
q
density (kg/m3)
r
standard deviation

Frg
m
OR
P
Pc
u
U
Usg
x
XLM

Subscripts
app
apparent
ave
average value
g
gas phase
l
liquid phase
max
maximum value
min
minimum value
tp
two-phase/wet gas

1b

gas densiometric Froude number ()


mass flow rate (kg/s)
over-reading ()
pressure (Pa)
confidence level (%)
standard uncertainty ()
relative standard uncertainty ()
superficial gas velocity (m/s)
gas mass fraction ()
Lockhart-Martinelli parameter ()

solution or more than one solution within a reasonable range [15].


Moreover, this method needs two throttle devices and two differential pressure sensors, thus increasing flow resistance and making
the metering system more complicated and costly.
The permanent pressure loss, which has contributed to on-line
detection of wet gas [16,17], can also be used as a characteristic
parameter to establish the auxiliary correlation. A recent study
carried out by Monni [18] applied the permanent pressure loss in
Venturi meter to derive flow rates of the gas and liquid phases.
However, a drawback of this method is that the permanent pressure loss or pressure loss ratio (the ratio of pressure loss to differential pressure) may not vary monotonously with flow rates in the
whole wet gas range [17]. Besides, although only one throttle
device is needed when using permanent pressure loss as the characteristic parameter, the metering system still consists of two differential pressure sensors, thus increasing the investment.
It is well-known that, due to the non-uniform distribution of
phases, turbulence, interaction between each phases and interaction between fluid and the pipe wall, flow parameters such as
pressure and differential pressure exhibit fluctuation when the
gas-liquid two-phase flow (wet gas) passes through the throttle
device [19]. In-depth study reveals that the fluctuation contains a
wealth of information relevant to the flow, and whereupon the
fluctuation can be considered as a characteristic parameter. The
differential pressure fluctuation has been applied in flow regime
identification [20,21] and aiding wet gas metering [15]. Shaban
[22] attempted to measure the gas and liquid flow rates by the
application of machine learning techniques to differential pressure
signals. Shen [23] managed to measure the gas mass fraction and
the total mass flow rate of air-water two-phase flow. Theoretically,
once the gas mass fraction and the total mass flow rate were
acquired by using differential pressure fluctuation [23], the gas
and liquid flow rates could be computed afterwards. Thus, flow
rate measurement of the two phases can be implemented by using
only one throttle device and one differential pressure sensor, making the metering system simplified and cost-effective.
Electrical parameters, such as the capacitance and the conductance, can act as characteristic parameters as well. Huang [24] used
a single-wire capacitance probe to measure the equivalent water
layer height in gas-liquid flow. Abbas [25] used a conductance
multiphase Venturi meter to measure the gas volume fraction at

the inlet and the throat of the Venturi. However, electrical parameters are easily affected by component, salinity and temperature.
When using differential pressure meters to measure the gas and
liquid flow rates in wet gas, a widely existed phenomenon is that
the prediction error of the gas phase is small while that of the liquid phase is large. Most researchers attempted to improve the
measurement accuracy by proposing new correlations [16,17,26].
However, there is no published work conducting error analysis
on differential pressure meters to explain this prediction error
pattern.
In this study, laboratory experiments are carried out to investigate the application of differential pressure fluctuation in wet gas
flow rate measurement. Afterwards, error analysis of metering
method based on differential pressure is concluded, aiming at
interpreting the prediction error pattern. Finally, characteristics
of the differential pressure signals are analyzed and their influences on the gas and liquid flow rates measurement are discussed.
2. Metering method based on differential pressure
When a differential pressure meter is used with wet gas flow,
the two-phase differential pressure from the meter is higher than
that, which would be read if the gas phase of the wet gas flowed
alone [1]. Consequently, the gas mass flow rate prediction due to
the two-phase differential pressure, which is generally termed
the apparent gas mass flow rate (mg,app) (indicated by Eq. (1)), is
larger than the actual gas mass flow rate (mg). Therefore, there is
a positive prediction error, usually called the over-reading (OR),
as is shown in Eq. (2).

mg;app EC d eAt
mg;app
OR
mg

q
2qg DPtp

1
2

where E is the velocity of approach; At is the minimum crosssectional area of a differential pressure meter, sometimes called
the throat; Cd is the discharge coefficient; e is the expansibility
coefficient; qg is the gas density; DPtp is the two-phase (wet gas)
differential pressure.
The over-reading denotes influences induced by the liquid
phase on the gas phase. Provided that the liquid flow rate or some

247

X. Zheng et al. / Experimental Thermal and Fluid Science 79 (2016) 245253

form of liquid-to-gas flow rate ratio is known, the liquid induced


positive error can be corrected by Eq. (3). Therefore, in order to
predict the gas flow rate, the over-reading needs to be calculated.
Experimental researches indicate that, for a set geometry meter,
the over-reading is dependent on the Lockhart-Martinelli parameter (XLM), the gas-to-liquid density ratio (qg/ql) and the gas densiometric Froude number (Frg), as expressed in Eq. (4). In this paper,
Eq. (4) is referred to as the over-reading correlation. The parameters, XLM and Frg are defined in Eqs. (5)(8) [1].

mg;app
OR
OR f X LM ; qg =ql ; Fr g
s
s
ml qg 1  x qg
X LM

x
mg ql
ql
mg
mg

x
mtp mg ml
s
qg
U sg
Fr g p
gD ql  qg

mg

U sg

3
4
5
6
7

mg
Aq g

where mtp is the total mass flow rate; ml is the liquid mass flow rate;
x is the gas mass fraction; Usg is the superficial gas velocity; A is the
area of the meter inlet.
Historically, a series of over-reading correlations have been proposed. The separated flow model [7] is the most basic model in wet
gas flow rate measurement with differential pressure meters. In
the separated flow model, the gasliquid two-phase flow is treated
separately as two single incompressible fluids flowing alone in the
pipe with identical discharge coefficients and differential pressures
to each other. The over-reading correlation derived from the separated flow model can be described as the simplest form:

OR 1 X LM

Many scholars had conducted modification to Eq. (9). For example, Murdock [3] assumed that the gas and liquid phases could
have individual flow equations with unique discharge coefficients,
and modified the correlation with experimental data; Bizon [4]
applied the similar method as Murdock used; Chisholm [5] considered the influence of shear between the gas and liquid phases; de
Leeuw [6] considered the influence of gas densiometric Froude
number (Frg); Lin [7] considered the influence of gas-to-liquid density ratio (qg/ql). The corresponding correlations are tabulated in
Table 1 and they are referred to as modified separated flow model
over-reading correlations in this paper.

If the liquid-to-gas flow rate ratio is known in prior, the overreading can be calculated by using the correlations in Table 1.
Then, the actual gas mass flow rate can be predicted by Eq. (10).
Afterwards, by incorporating Eqs. (6) and (10), equations for predicting the liquid mass flow rate (ml) and the total mass flow rate
(mtp) can be rearranged as Eqs. (11) and (12).

mg

EC d eAt

q
2qg DPtp

10

OR
q

EC d eAt 2qg DP tp 1  x

ml
OR
q x
mtp

EC d eAt

11

2qg DPtp

12

OR  x

Eqs. (10)(12) indicate that, in order to implement the flow rate


measurement, the gas mass fraction x (or some form of gas-toliquid flow rate ratio information) needs to be supplied. A feasible
approach is to establish an auxiliary correlation to calculate the gas
mass fraction. The key issue in establishing the auxiliary correlation is the selection of flow characteristic parameters. In this study,
the fluctuation of differential pressure signals is chosen as the
characteristic parameter. The relationship between the differential
pressure fluctuation and the gas mass fraction is obtained through
experiments, and then the specific form of the auxiliary correlation
is obtained by fitting experimental data.
3. Experimental facilities and scheme
A conical orifice plate with flange taps, exhibited in Fig. 1(a), is
used in the experiment, where the diameter of the throat is 15 mm,
the equivalent diameter ratio is 0.3, the inlet-cone angle is 60, the
outlet-cone angle is 70, and the width of the platform connecting
the inlet cone and the outlet cone is 1.5 mm. As is shown in Fig. 1
(b), the test section consists of two straight pipes with the inner
diameter of 50 mm, a conical orifice plate, two connecting pipes,
two needle values, a pressure transducer and a differential pressure transducer.
The experimental setup shown in Fig. 2 consists of the air and
the water loops, with whom flow meters are equipped to measure
the single-phase flow rate. There is a transparent observation section made of quartz installed before the orifice palate section. The
diameter of the quartz tube is 50 mm. A high speed camera is used
to record the flow.
The flow system pressure ranges from 0.7 MPa to 1.0 MPa. The
gas mass flow rate is between 200 kg/h and 400 kg/h while the liquid mass flow rate is between 7 kg/h and 390 kg/h, so that the gas
mass fraction ranges from 0.4 to 0.97 while the Lockhart-Martinelli

Table 1
Typical over-reading correlations.
Model

Throttle element

Correlation

Separated flow model

Orifice plate

OR = 1 + XLM

Murdock [3]

Orifice plate

OR = 1 + 1.26XLM

Bizon [4]

Orifice plate

OR = a + bXLM

Venturi
Chisholm [5]

Orifice plate

OR = a + bXLM
q
OR 1 C Ch X LM X 2LM

 14  14
q
q
C Ch qg q l
l
g

de Leeuw [6]

Venturi

OR

q
1 C de X LM X 2LM

q n  n
C de qg qql
l
g

Lin [7]

Orifice plate

OR = 1 + hLXLM

a 1:0372 b 1:0179 b 0:45


a 1:0818 b 0:9999 b 0:70
a 1:0426 b 1:0799 b 0:58

n 0:41;
n 0:6061  e0:746Frg ;

 
 2
q
q
hL 1:48625  9:26541 qg 44:6954 qg
l
l
 3
 5
qg
qg 4
q
60:615 q
 5:12966 q 26:5743 qg
l

0:5 6 Frg 6 1:5


Fr g P 1:5

248

X. Zheng et al. / Experimental Thermal and Fluid Science 79 (2016) 245253

Fig. 1. Test section.

Fig. 2. Experimental setup: 1-air compressor, 2-air tank, 3-air filter, 4-air freezing dryer, 5-ball valve, 6-pressure gauge, 7-thermometer, 8-globe valve, 9-air mass flow meter,
10-check valve, 11-control valve, 12-water tank, 13-sluice valve, 14-centrifugal pump, 15-electromagnetic flow meter, 16-gas-liquid mixer, 17-temperature transducer,
18-pressure transducer, 19-differential pressure transducer, 20-gas-liquid separator, 21-high speed camera.

parameter ranges from 0.003 to 0.170. As can be seen from the


photographs taken by the high speed camera (Fig. 3), the flow patterns of the experimental conditions are stratified flow, which is
consistent with the flow distribution in the Mandhane flow pattern
map [27]. The uncertainties of the temperature, the pressure, the
differential pressure, the gas mass flow rate and the liquid mass
flow rate are 1.07%, 0.37%, 0.22%, 0.94% and 4.26%, respectively.
The experimental data are collected with the NI USB-6229 data
acquisition module and the LabVIEW based software. The sampling
frequency is 50 Hz and the sampling time is 60 s. For every experimental point, we remain the conditions unchanged (i.e. with the
same pressure, temperature, gas flow rate and liquid flow rate)
and collect three data sets at different time.
4. Experimental results and analysis
Differential pressure signals exhibit fluctuation when the gasliquid two-phase flow passes through the throttle device. The
dimensionless parameter B is used to reflect the differential pressure fluctuation:

rDPtp
DPtp

13

where DP tp is the mean value of the differential pressure signals; r


(DPtp) is the standard deviation of the differential pressure signals.
As illustrated in Fig. 4, B decreases as the gas mass fraction x
increases, and the corresponding correlation, namely the auxiliary
correlation, is fitted as Eq. (14):

x e0:100835:57B

14

The gas mass fraction can be calculated by Eq. (14), and the prediction result is shown in Fig. 5. Then, the gas mass flow rate, the
liquid mass flow rate and the total mass flow rate can be obtained
by substituting the over-reading OR and the gas mass fraction x
into Eqs. (10)(12). In this paper, the separated flow model, the
Murdock correlation, the Chisholm correlation and the Lin correlation are used to calculate the over-reading OR. The prediction
results of the flow rates are shown in Figs. 68.
4.1. Error propagation rules of metering method based on differential
pressure
Figs. 68 indicate that the prediction results of different overreading correlations have minor difference from each other and
they have same error pattern: the prediction error of the gas phase
is small, while the prediction error of the liquid phase is large; the

X. Zheng et al. / Experimental Thermal and Fluid Science 79 (2016) 245253

249

Fig. 6. Comparison between experimental and estimated gas mass flow rate.

Fig. 3. Flow pattern of experimental conditions.

Fig. 7. Prediction error of liquid mass flow rate.

Fig. 4. Relationship between B and x.

Fig. 5. Prediction error of gas mass fraction.

Fig. 8. Prediction error of total mass flow rate.

250

X. Zheng et al. / Experimental Thermal and Fluid Science 79 (2016) 245253

As to low pressure laboratory experiments, the operating pressure


is usually lower than 3 MPa, which gives:

Table 2
Comparison of prediction errors (Pc is the confidence level).
Author

Zhang [14]
Monni [18]
Present study

Relative error (%)

C g < 0:5

mg

ml

mtp

7%
5%
10%
(Pc = 87%)

5%
2%
5%
(Pc = 94%)

12%
30%
25%
(Pc = 74%)

10%

10%
(Pc = 90%)

prediction error of the total mass flow rate is equivalent to that of


the gas mass fraction. Besides, we compared our results with
Zhang [14] and Monni [18]. The comparison in Table 2 shows that
previous work has the same error pattern as in the present study. It
is worth noting that both previous methods and our method are
founded on the basis of over-reading correlations, but with different characteristic parameters used for calculating the gas mass
fraction. Accordingly, it is conceivable to infer that the same error
pattern results from the over-reading correlations.
In the following part, the error propagation rules of metering
method based on differential pressure are investigated, provided
that the gas mass fraction has been supplied in prior via the auxiliary correlation. As stated previously in Section 2, the other correlations in Table 1 can be regarded as modified separated flow
model over-reading correlations, and Figs. 68 also show that the
separated flow model has the same error pattern as the others.
Therefore, the following error analysis is based on the separated
flow model over-reading correlation (i.e. Eq. (9)) for simplicity. It
is rational to claim that the results of error analysis apply to all
the over-reading correlations developed from separated flow
model.
According to error propagation theory, the deviation of indirectly measured parameter y can be estimated by that of the
directly measured parameters xi:

dy

n
X
@f
dxi
@x
i
i1

15

where y = f (x1, x2, . . ., xn), d denotes the deviation.


Substituting Eqs. (5) and (9) into Eq. (10) gives the equation for
the gas mass flow rate:

q
mg EC d eAt 2qg DP tp

q
x 1  x qg =ql

16

@m

g
dx
dmg
@x

mg
mg


1
1  x q
1
qg =ql
x

q
qg =ql dx
dx
 Cg
x
x
x

17

where Cg is the error transfer coefficient for gas mass flow rate.


1
1  x q
Cg 1
qg =ql
x
q
qg =ql
<
x

q
qg =ql
x


1 X LM

q
qg =ql
x

18

Noting that XLM 6 0.3 for wet gas, and

1  x q

qg =ql 6 0:3 )
x

q
qg =ql

6 0:3

q
qg =ql

Eq. (18) can then be rearranged to C g < 0:3

q
qg =ql .

X LM

In light of the analysis above, it is clear that the relative error of


the gas mass flow rate is much smaller than that of the gas mass
fraction.
Substituting Eqs. (5) and (9) into Eq. (11) gives the equation for
the liquid mass flow rate:

ml EC d eAt

19

Taking air-water two-phase flow for instance, Cg is less than 1 as


q
long as the pressure is no more than 37 MPa (where qg =ql < 0:7).

q
2qg DPtp

1x
q
x 1  x qg =ql

21

then, the relative error of the liquid mass flow rate can be estimated
in the same way:


1
@ml
dx
dml
1  x q
1
dx
dx
 C l
@x
 1
qg =ql
x
1x x
x
ml
ml

22

where Cl is the error transfer coefficient for liquid mass flow rate:


1
1  x q
1
1
1

Cl 1
qg =ql

x
1  x 1 X LM 1  x

23

Likewise, XLM 6 0.3 for wet gas, rearranging Eq. (23) gives

Cl P

1
1:31  x

24

It is worth noting that, for wet gas with x > 0.23, Cl is larger than
1, which means that the relative error of liquid mass flow rate is
amplified due to error propagation. Moreover, the greater the gas
mass fraction, the higher the degree of amplification. When
x = 0.99, for instance, the relative error of liquid mass flow rate is
77 times larger than that of the gas mass fraction.
Substituting Eqs. (5) and (9) into Eq. (12) gives the equation for
the total mass flow rate:

q
2qg DPtp
q
ml
x 1  x qg =ql
EC d eAt

25

Then, the relative error of the total mass flow rate can be estimated similarly:
@mtp
@x

Therefore, the relative error of the gas mass flow rate can be
estimated by that of the gas mass fraction:

20

dx

dmtp

mtp
mtp


q
1  qg =ql x
dx
dx
q  C tp
 q 
x
qg =ql 1  qg =ql x x

26

where Ctp is the error transfer coefficient for total mass flow rate.
It is evident that Ctp is less than but close to 1, so the relative
error of the total mass flow rate is smaller but equivalent to that
of the gas mass fraction. In addition, the minus sign before the
error transfer coefficient in Eq. (26) denotes that the error tendency of total mass flow rate is opposite to that of gas mass fraction (see Figs. 5 and 8).
To sum up, due to error propagation, the relative error of the gas
mass flow rate is much smaller than that of the gas mass fraction,
whereas the relative error of the liquid mass flow rate is much larger than that of the gas mass fraction; the relative error of the total
mass flow rate is equivalent to that of the gas mass fraction.
Besides, the error pattern is a universal phenomenon for wet gas
metering method based on differential pressure.
The physical essence of the great gap between measurement
accuracy of the gas and liquid phases lies in the fact that the two
phases have exerted different influence on the two-phase differential pressure. According to the separated flow model [7], the mass
flow rate equations for the gas and liquid phases if they flowed
alone would be:

X. Zheng et al. / Experimental Thermal and Fluid Science 79 (2016) 245253

q
mg EC d eAt 2qg DPg
p
ml EC d eAt 2ql DPl

251

27
28

where DPg and DPl are the pressure drop if gas or liquid flowed
alone, respectively. Then, the single phase differential pressure ratio
can be deduced from Eqs. (27) and (28):

 2
DP l
ml qg

X 2LM
DP g
mg ql

29

Considering that XLM is less than 0.3 for wet gas by definition,
the single phase differential pressure ratio is less than 0.09, which
means that, compared with the gas phase, the liquid phase has
made a much smaller contribution to the two-phase differential
pressure DPtp. As a consequence, the two-phase differential pressure is mainly influenced by the predominant gas phase and is
insensitive to the liquid phase.
Fig. 10. Variation of differential pressure mean value.

4.2. Characteristics of differential pressure signals


Revisiting of the relative errors shown in Table 2 indicates that
the measurement accuracy in the present study is lower than previous work. These methods are all based on over-reading correlations but with different characteristic parameters used for
establishing the auxiliary correlation. Zhang [14] adopts the across
throttle device pressure drop, while Monni [18] adopts the permanent pressure loss. These characteristic parameters are both differential pressure in essence, but with different low pressure ports.
Zhang [14] and Monni [18] used the mean value of the differential
pressure signals. In the present study, the fluctuation of the differential pressure signals is used. It is the characteristics of differential pressure fluctuation that bring down the measurement
accuracy. In this section, the mean value and the fluctuation of
the differential pressure signals are investigated systematically.
Fig. 9 shows that when the pressure is fixed, the differential
pressure mean value increases with both the gas and liquid flow
rates, which means that the differential pressure mean value can
reflect flow rate changes of wet gas. However, as claimed previously, influence of the liquid phase on the differential pressure
mean value is small compared to that of the gas phase.
Data in Fig. 10 are collected from experiments conducted on different dates with the same pressure and gas mass flow rate. As can
be seen, the differential pressure mean values show good consis-

Fig. 9. Relationship between differential pressure mean value and flow rates.

tency, and the relative deviation from the fitted line is less than
0.46%. Conclusions are then drawn that the differential pressure
mean value has good repeatability and can therefore perform as
a robust characteristic parameter of the two-phase flow.
As a common practice, the standard deviation is used to quantify the degree of fluctuation. Fig. 11 shows that the differential
pressure standard deviation increases with the gas and liquid flow
rates for most cases. This tendency indicates that the fluctuation of
differential pressure signals can reflect flow rate changes of wet
gas, and therefore it is considered as a characteristic parameter
of the two-phase flow.
In this paper, three sets of differential pressure signals are collected under each experimental condition, for the purpose of evaluating the repeatability of the differential pressure fluctuation.
Consequently, three differential pressure standard deviations are
calculated for each operating point. As can be seen from Fig. 12
(a), the average value of the three differential pressure standard
deviations (denoted by circular dots), increases with the liquid
mass flow rate, providing that the pressure and the gas mass flow
rate remain unchanged. However, the variation of the differential
pressure standard deviation from the average value is much wider
than that of the differential pressure mean value, with the maximum relative deviation beyond 10%, as shown in Fig. 12(b).
(The error bar in Fig. 12(a) denotes the difference between the

Fig. 11. Relationship between differential pressure standard deviation and flow
rates.

252

X. Zheng et al. / Experimental Thermal and Fluid Science 79 (2016) 245253

operating condition changes, meaning that the value of U[r(DPtp)]


is stochastic. The reason for this phenomenon is that gas-liquid
two-phase flow is a stochastic flow in nature [19].
In the present study, the gas mass fraction was calculated using
Eq. (14). Thus, according to uncertainty theory, the relative standard uncertainty of the gas mass fraction (Ux) can be estimated
by that of the differential pressure standard deviation:

Ux

maximum and the minimum values of the three differential pressure standard deviations.) The wide variation range means that the
repeatability of the differential pressure standard deviation is poor.
Herein, the relative standard uncertainty of the differential pressure standard deviation, defined by Eq. (30), is used to quantify
the repeatability:

urDPtp 
rDPtp av e

30

where [r(DPtp)]ave is the average value of the three differential pressure standard deviations; u[r(DPtp)] is the standard uncertainty of
differential pressure standard deviation. Due to the small number
of samples, u[r(DPtp)] is estimated by Eq. (31):

urDPtp 

rDPtp max  rDPtp min


dn

35:57B  UrDPtp 

32

In present cases, the relative standard uncertainty of the gas


mass fraction ranges from 0.04% to 16.67%, which means that the
poor repeatability of the differential pressure standard deviation
leads to considerable prediction error of the gas mass fraction.
Therefore, due to the poor repeatability, the differential pressure
fluctuation may be suitable for qualitative analysis (e.g. flow
regime identification) rather than quantitative calculation (i.e. flow
rate measurement).

Fig. 12. Variation of differential pressure standard deviation.

UrDPtp 

ux


2 
2
@x
u
r

D
P

tp
@rDPtp 

31

where [r(DPtp)]max and [r(DPtp)]min are the maximum and minimum values of the three differential pressure standard deviations,
respectively; dn is the coefficient (d3 = 1.693, for n = 3).
The variation of U[r(DPtp)] with the liquid mass flow rate under
different pressure and gas mass flow rate is presented in Fig. 13.
The results illustrate that U[r(DPtp)] varies between 0.3% and
18.5% for all conditions. Besides, U[r(DPtp)] shows no rules as the

5. Conclusions
In this paper, the error analysis of gas and liquid flow rates
metering method based on differential pressure in wet gas is investigated by theoretical analysis and laboratory experiments. The
error propagation analysis demonstrates that metering methods
based on differential pressure will exhibit similar error pattern.
The prediction error of the gas mass flow rate is small, while that
of the liquid mass flow rate is large. In addition, the higher the
gas mass fraction of the wet gas is, the larger the liquid prediction
error will be. This is because the two-phase differential pressure in
wet gas mainly depends on the gas phase and is insensitive to the
liquid phase. Investigation on characteristics of the differential
pressure signals reveals that both the mean value and the fluctuation of the signals can be taken as characteristic parameters.
However, these two statistics parameters show different properties. The mean value of the differential pressure signals can reflect
flow rate changes in wet gas and has good repeatability. In the
experiments, the relative deviation of the differential pressure
mean value is less than 0.46%. Although the fluctuation of the differential pressure signals increases with the gas and liquid flow
rates, the repeatability of the fluctuation is poor. In present cases,
the relative standard uncertainty of the differential pressure fluctuation is from 0.3% to 18.5%, thus leading to the relative standard

Fig. 13. Variation of U[r(DPtp)] under different conditions (a) P = 0.7 MPa and (b) mg = 300 kg/h.

X. Zheng et al. / Experimental Thermal and Fluid Science 79 (2016) 245253

uncertainty of the calculated gas mass fraction from 0.04% to


16.67%. Owing to poor repeatability and error propagation, metering method based on the fluctuation of the differential pressure
signals will result in appreciable liquid phase prediction error,
especially for wet gas with high gas mass fraction.
Acknowledgements
This study was supported by National Natural Science
Foundation of China under Grant No. 51276140 and by National
Key Scientific Instrument and Equipment Development Project,
China under Grant No. 51527808.
References
[1] ASME, Wet Gas Flowmetering Guideline. ASME MFC-19G-2008, ASME, New
York, USA, 2008.
[2] ISO, ISO/TR 11583-2012, Measurement of Wet Gas Flow by Means of Pressure
Differential Devices Inserted in Circular Cross-Section Conduits, International
Standards Organization, 2012.
[3] J.W. Murdock, Two-phase flow measurements with orifices, J. Basic Eng. 84
(1962) 419433.
[4] Z.H. Lin, S.Z. Wang, D. Wang, GasLiquid Two-Phase Flow and Boiling Heat
Transfer, Xian Jiaotong University Press, Xian, 2003, pp. 169186 (in Chinese).
[5] D. Chisholm, Research note: two-phase flow through sharp-edged orifices, J.
Mech. Eng. Sci. 19 (1977) 128130.
[6] R. De Leeuw, Liquid correction of Venturi meter readings in wet gas flow, North
Sea Flow Measurement Workshop 1997, Norway, 1997.
[7] Z.H. Lin, Two-phase flow measurements with sharp-edged orifices, Int. J.
Multiphase Flow 8 (1982) 683693.
[8] R. Steven, Liquid property and diameter effects on Venturi meters used with
wet gas flows, in: International Fluid Flow Measurement Symposium, 2006.
[9] Michael Reader-Harris, An improved model for Venturi-tube overreading in
wet gas, in: 27th International North Sea Flow Measurement Workshop, 2009.
[10] D.H. He, B.F. Bai, Y. Xu, X. Li, A new model for the V-Cone meter in low pressure
wet gas metering, Meas. Sci. Technol. 23 (2012) 125305. 9pp.
[11] D.H. He, B.F. Bai, Two-phase mass flow coefficient of V-Cone throttle device,
Exp. Therm. Fluid Sci. 57 (2014) 7785.

253

[12] Y. Xu, L. Gao, Y. Zhao, H.X. Wang, Wet gas overreading characteristics of a longthroat Venturi at high pressure based on CFD, Flow Meas. Instrum. 40 (2014)
247255.
[13] Y. Xu, Q. Zhang, T. Zhang, X.L. Ba, An overreading model for nonstandard
Venturi meters based on H correction factor, Measurement 61 (2015) 100
106.
[14] F.S. Zhang, F. Dong, C. Tan, High GVF and low pressure gasliquid two-phase
flow measurement based on dual-cone flowmeter, Flow Meas. Instrum. 21
(2010) 410417.
[15] L.J. Xu, J. Xu, F. Dong, T. Zhang, On fluctuation of the dynamic differential
pressure signal of Venturi meter for wet gas metering, Flow Meas. Instrum. 14
(2003) 211217.
[16] Y. Xu, C. Yuan, Z.H. Long, Q. Zhang, T. Zhang, Research the wet gas flow
measurement based on dual-throttle device, Flow Meas. Instrum. 34 (2013)
6875.
[17] D.H. He, B.F. Bai, J. Zhang, X.W. Wang, Online measurement of gas and liquid
flow rate in wet gas through one V-Cone throttle device, Exp. Therm. Fluid Sci.
75 (2016) 129136.
[18] G. Monni, M. De Salve, B. Panella, Two-phase flow measurements at high void
fraction by a Venturi meter, Prog. Nucl. Energy 77 (2014) 167175.
[19] L.J. Guo, Dynamics of Two-Phase and Multi-Phase Flow, Xian Jiaotong
University Press, Xian, 2002 (in Chinese).
[20] G. Matsui, Identification of flow regimes in Vertical gas-liquid two-phase flow
using differential pressure fluctuations, Int. J. Multiphase Flow 10 (1984) 711
719.
[21] B.F. Bai, L.J. Guo, L. Zhao, Identification of flow pattern regimes in vertical
upward steam-water two phase flow using differential pressure fluctuations, J.
Chem. Ind. Eng. (China) 50 (1999) 799805.
[22] H. Shaban, S. Tavoularis, Measurement of gas and liquid flow rates in twophase pipe flows by the application of machine learning techniques to
differential pressure signals, Int. J. Multiphase Flow 67 (2014) 106117.
[23] G.Q. Shen, Z.H. Lin, A dynamic method for dual parameters of gasliquid
biphase flow, Acta Metrol. Sin. 14 (1993) 140145 (in Chinese).
[24] S.F. Huang, X.G. Zhang, D. Wang, Z.H. Lin, Equivalent water layer height
(EWLH) measurement by a single-wire capacitance probe in gasliquid flows,
Int. J. Multiphase Flow 34 (2008) 809818.
[25] Abbas H.A.M. Hasan, G.P. Lucas, Experimental and theoretical study of the gas
water two phase flow through a conductance multiphase Venturi meter in
vertical annular (wet gas) flow, Nucl. Eng. Des. 241 (2011) 19982005.
[26] Richard Steven, Andrew Hall, Orifice plate meter wet gas flow performance,
Flow Meas. Instrum. 20 (2009) 141151.
[27] J.M. Mandhane, G.A. Gregory, K. Aziz, A flow pattern map for gasliquid flow in
horizontal pipes, Int. J. Multiphase Flow 1 (1974) 537553.