Escolar Documentos
Profissional Documentos
Cultura Documentos
log Re : 1
Britton and Stark (1986) conducted experiments to determine dis-
charge coefcients of sharp edged orice plates which are hydraulical-
ly similar to the standard orice plate. Reynolds numbers in the tests
ranged from 1000 to 50,000 with the corresponding discharge coef-
cients varying from 0.605 to 0.695 for the sharp edged orice plate.
Singh et al. (2006) studied the discharge coefcients of the V-Cone
ow meter. They proposed that for Reynolds numbers between 1250
and 254,000 that the average discharge coefcient for the beta values
of 0.64 and 0.77 were 0.7256 and 0.7310, respectively.
A study by Tan et al. (2009) was conducted to determine the dis-
charge coefcients of two-phase ow using a V-cone ow meter. The
study found that the Reynolds numbers ranging from 25,000 to
85,000 had a discharge coefcients of 0.80.85 respectively, with an
average value of 0.83.
Banchhor et al. (2002) explored the effects of vertex angles and
the apex height/diameter ratio (H/D) on the discharge coefcient of
wedge ow meters using CFD. It was found that as the wedge vertex
radii was increased from 0 to 6 mm the resultant C value also in-
creased and the C remained constant for much of the Re range tested.
A study conducted by Yoon et al. (2007) resulted in a proposed
discharge coefcient equation for segmental wedge ow meters.
Five different beta ratios were tested and had varying discharge coef-
cients, but they formulated Eq. (2) to predict the constant discharge
coefcients for segment height to diameter ratios (H/D) values rang-
ing from 0.3 to 0.7.
C
d
0:9136
H=D
0:1303
23:363
H=D
4
50:378
H=D
3
37:246
H=D
2
: 2
3. Methodology
A review of the literature demonstrated that there is little infor-
mation pertaining to low Reynolds number discharge coefcient per-
formance for several different designs of differential producing ow
meters. Consequently, this study was performed so that users of
such devices may understand the discharge coefcient characteristics
at low Reynolds numbers.
The steady, incompressible, Reynolds-averaged NavierStokes
equations were solved using the general purpose computational
uid dynamics solver FLUENT (ANSYS Inc., 2009). Pressurevelocity
coupling was accomplished using the SIMPLE procedure (ANSYS
Inc., 2009). Interpolation of variables to cell faces was performed
using a second-order upwinding scheme, whereas diffusion terms
were discretized using second-order central differencing. For calcula-
tions completed at pipe Reynolds numbers greater than 2300 (based
on hydraulic diameter and mean velocity), a realizable k-epsilon
model was used for turbulence closure (ANSYS Inc., 2009). At lower
Reynolds numbers the ows were assumed laminar and hence no
turbulence model was implemented. The iterative convergence cri-
teria were satised when the normalized residuals for all equations
were decreased by a minimum of six orders of magnitude. Observa-
tions indicated that further reductions in residuals resulted in no dis-
cernible differences in the results.
An unstructured computational mesh was created for each of the
thirteen numerical models examined. Meshes were converted from
tetrahedral cells to polyhedra cells before beginning analysis. Typical
meshed geometries contained on the order of 400,000 polyhedra
cells. It should be noted that the near-wall mesh was developed
for implementation of wall functions with the turbulence model. A
V-cone mesh shown in Fig. 2 is representative of the mesh densities
of the other ow meters investigated. Fully developed pipe ow
was used to develop k and epsilon levels at the velocity inlet bound-
ary condition. The pressure outlet boundary condition used typical
downstream operating pressures.
Surface planes were created to determine resultant pressures at
each of the ow meter's pressure tap locations. The planes were cross
sectional views located at both of the tap locations for each model ex-
cept for the V-cone ow meter. Pressures for the Venturi and orice
plate meters were averaged about the perimeter of the surface planes
at standard locations. The wedge meter pressures were averaged from
Fig. 1. Sketches of meters.
560 C.L. Hollingshead et al. / Journal of Petroleum Science and Engineering 78 (2011) 559566
3 perimeter point locations nearest the top of the meter as was the up-
streampressure for the V-cone meter, while the downstreamwas aver-
aged about the perimeter. The V-cone models were treated differently,
because the downstream pressure reading was the average pressure of
the multiple pressure inlets on the backside of the cone.
The following equation was used to determine the mass ow rate
through the differential pressure ow meters (Miller, 1996):
q N
Cd
2
1 d=D
4
_
_
_
_
_
_
_ Yf
h
w
_
3
where, q is the discharge, N is a dimensional constant, C is the dis-
charge coefcient of the ow meter, d is the effective throat diameter,
Yf() is the thermal correction factor, D is the inlet diameter of the
meter, and hw is the differential pressure between the pressure
taps. Miller's general equation is commonly used and widely accepted
in the ow meter industry.
The following equations were used to determine the value of beta
(), where is dened as the ratio of the owmeter's effective throat
diameter to the meters inlet diameter, for the different ow meters
investigated. In the case of the Venturi and orice meters, the effec-
tive diameter is the actual throat diameter. For the wedge meter,
is dened as:
wedge
1
a cos 1
2H
D
_ _
2 1
2H
D
_ _
H
D
H
D
_ _
2
_ _
0:5
_ _ _ _
0:5
: 4
Here, H is the wedge segment height, and D is the inlet diameter
upstream of the wedge (PFSI, Primary Flow Signal Inc., 2001). For
the Venturi and orice meters, the values of are dened as:
d=D 5
where d is the throat diameter, and D is the inlet diameter. The rela-
tion for the V-cone meter is given as:
Vcone
D
2
I
d
2
cone
_
=D
I
6
where D
I
is the inlet diameter, and d
cone
is the diameter of the cone at
its largest cross sectional area. Finally, the effective diameter for any
of the given ow meters can be dened as
d D
I
7
where the effective diameter is dened as the diameter of a circular
duct that has the same area as the throat of interest.
4. Results and discussion
The results are presented primarily in the form of line plots of dis-
charge coefcient vs. Reynolds number on semi-log graphs. In partic-
ular, results for the Venturi, orice, V-cone and wedge meters are
shown in Figs. 36, respectively. Experimental data using water is
also provided for comparative purposes in each of the gures.
Heavy oil was used for ows where Re20,000, while water was
used to obtain results at higher Reynolds numbers. As a test of consis-
tency, both water and heavy oil were run at a Re of 20,000 for all of
the models. The results revealed that for each case, discharge coef-
cients resulting from the CFD calculations were within 0.2% from
one another, and within 3% of the lab data. Table 1 provides a summa-
ry of all models tested.
Three different Venturi models were used to acquire the discharge
coefcient data shown in Fig. 3. Two of the Venturi meters had an
inlet diameter (D) of 15.405 cm, and a value of 0.661. The primary
difference between the two Venturi meters was their respective
inlet geometries. In one case the inlet smoothly transitioned to the
throat using a parabolic cone. The other consisted of a segmented
cone, having two break points as it reduced from the inlet to the
throat. The third Venturi also consisted of a segmented cone with
an inlet diameter of 15.405 cm, and a value of 0.242. The results in-
dicate that the discharge coefcient remains constant, (varying less
than 1%) for 75,000Re50,000,000. As the Reynolds numbers
were reduced from 30,000 to 1, the discharge coefcient values
smoothly trended from approximately 0.96 to approximately 0.15.
Fig. 2. V-cone polyhedra mesh.
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
1 10 100 1,000 10,000 100,000 1,000,000 10,000,000
D
i
s
c
h
a
r
g
e
C
o
e
f
f
i
c
i
e
n
t
Reynolds Number
Sharp Beta = 0.242 Sharp Beta = 0.242 Lab Data
Sharp Beta = 0.661 Sharp Beta = 0.661 Lab Data
Smooth Beta = 0.661 Smooth Beta = 0.661 Lab Data
Fig. 3. Venturi meter discharge coefcients.
561 C.L. Hollingshead et al. / Journal of Petroleum Science and Engineering 78 (2011) 559566
The data for the Venturi meter was compared to results presented in
(Miller et al., 2009) where it was reported that for Re75,000 the dis-
charge coefcients began to trend downward from their constant co-
efcients. The physical laboratory data presented in Fig. 3 ranged
from 3980Re506,000, which demonstrates the limited range of
physical data.
In terms of the orice plate meters, four standard models with dif-
ferent values of were examined to better understand the vs. dis-
charge coefcient relationship with varying Reynolds numbers.
Values of equal to 0.50, 0.60, 0.65, and 0.70 were used to model
the orice plates. Orice meter inlet diameters ranged from 15.41 to
20.27 cm. (The purpose of the different diameters was to determine
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1 10 100 1,000 10,000 100,000 1,000,000 10,000,000
D
i
s
c
h
a
r
g
e
C
o
e
f
f
i
c
i
e
n
t
Reynolds Number
Beta = 0.50 Beta = 0.60
Beta = 0.65 Beta = 0.70
Beta = 0.50 Miller, 1996 Beta = 0.60 Miller, 1996
Beta = 0.65 Miller, 1996 Beta = 0.70 Miller, 1996
Fig. 4. Orice plate discharge coefcients.
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
1 10 100 1,000 10,000 100,000 1,000,000 10,000,000
D
i
s
c
h
a
r
g
e
C
o
e
f
f
i
c
i
e
n
t
Reynolds Number
Beta = 0.611 Beta = 0.611 Lab Data
20.3 cm Beta = 0.5023 20.3 cm Beta = 0.5023 Lab Data
15.2 cm Beta = 0.5023 15.2 cm Beta = 0.5023 Lab Data
Fig. 5. Wedge ow meter discharge coefcients.
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
Reynolds Number
D
i
s
c
h
a
r
g
e
C
o
e
f
f
i
c
i
e
n
t
Beta = 0.6611 Beta = 0.6611 Lab Data
Beta = 0.6995 Beta = 0.6995 Lab Data
Beta = 0.8203 Beta = 0.8203 Lab Data
10 100 1,000 10,000 100,000 1,000,000 10,000,000 100,000,000
Fig. 6. V-cone meter discharge coefcients.
562 C.L. Hollingshead et al. / Journal of Petroleum Science and Engineering 78 (2011) 559566
if results were dependent on the inlet diameter.) The numerical re-
sults, along with Miller's data that ranged from 5Re201,000, are
shown in Fig. 4 (Miller, 1996). The results indicate that the discharge
coefcient responded somewhat differently than the other meters
over the approximate range 100Re10,000. In particular, the dis-
charge coefcients do not decrease in monotonic fashion, rather
they exhibit a local maximum at Re300. This may be attributed
to the effect of the orice plate on the downstream velocity prole
(Britton and Stark, 1986). For Reynolds numbers trending from 100
down to 1, the value of the discharge coefcient decreases from ap-
proximately 0.70.8 to approximately 0.20.25 depending on the
value of for a given meter.
Three wedge ow meter models were investigated in order to de-
termine their discharge coefcient data over a wide range of Reynolds
numbers. Two of the three models had values of =0.5023 (H/
D=0.3), but different diameters. The purpose of this was to deter-
mine if results were dependent on pipe diameter. The other value of
tested was 0.611(H/D=0.4) for a meter with an inlet diameter of
20.27-centimeter. Resulting numerical discharge coefcients, and
physical laboratory data ranging from 4300Re116,100, are
shown in Fig. 5. The results reveal little difference between the two
models that had identical values of , but different inlet diameters.
These discharge coefcients had a variance of less than 1% over the
range 500Re50,000,000. The larger values resulted in a de-
crease in the constant discharge coefcient. This is due to the fact
that as increases, so too does the ratio H/D. Computational uid dy-
namics research performed by Banchhor in 2002 and research in the
current study were found to be in agreement within 3% for the con-
stant discharge coefcient of H/D=0.4 (Banchhor et al., 2002).
The V-cone ow meter was modeled using three different
values. The results revealed that as the value of increased from
0.66 to 0.82, the corresponding discharge coefcient values decreased
from approximately 0.80 to 0.73. For the constant discharge coef-
cient the V-cone meter deviations appeared to be similar to that of
the Venturi meter with a variance of less than 1% for the Reynolds
number range of 30,00050,000,000. Resulting numerical discharge
coefcient data, and physical laboratory data ranging from
37,400Re955,700, are shown in Fig. 6. The V-cone ow meter
data was compared to previous physical data presented by Tan (Tan
et al., 2009), where oil or an emulsion mixture of oil and water
were used in the experiments. Tan studied the discharge coefcient
in the Reynolds number ranges between 25,000 and 80,000, and the
results showed agreement within 5% of the current study (Tan et al.,
2009). Tables 47 represent the numerical values from Figs. 36,
respectively.
Compared in Table 2 are the calculated ow rates with and with-
out differing discharge coefcients for ow meters that are often as-
sumed to have constant discharge coefcients. The rapid increase in
difference between the two calculated ow rates is clearly shown as
the Reynolds number decreases for the Venturi, V-cone, and wedge
meters.
The average trends for the discharge coefcient deviation fromthe
constant discharge coefcient are given in Fig. 7. The gure shows the
relative change of the average discharge coefcient fromaverage con-
stant coefcient for each of the ow meter types in the study. For ex-
ample, if a V-Cone owmeter had an inlet Reynolds number of 10, the
value of the discharge coefcient relative to the constant discharge
coefcient is 0.25.
As shown in Fig. 7 each of the meter types exhibits a discharge co-
efcient reduction as the Reynolds number decreases, the exception
is the orice meter, where the discharge coefcient increases as the
Reynolds number decreases to 200. Table 2 demonstrates that each
of the meters discussed has a different (lower) Reynolds number
limit to where the assumption of a constant discharge coefcient re-
mains valid. The discharge coefcient for the wedge ow meter
remained most consistent, varying less than 1% for ows above a
Reynolds number of 500. At the other extreme, the V-cone ow
meter has a discharge coefcient change of over 1120% at the lowest
Reynolds numbers relative to the constant discharge coefcient.
To assist users of the data presented in Fig. 7, equations were de-
veloped to describe the average trends of each of the four meter
types investigated. Curve tting software was used to identify equa-
tions that were well behaved over the Reynolds number range inves-
tigated and that had very strong correlation to the data collected. The
equations take the form
C C
8
where C
is the con-
stant coefcient of discharge for the corresponding ow meter. Aver-
age correction coefcient equations for each type of ow meter are
given in Eqs. (9)(12) and the coefcients for each equation are
given in Table 8 along with the coefcient of determination or R
2
Table 1
Summary of ow meters.
Type Diameter (cm) Beta value Type Diameter (cm) Beta value
Venturi 15.41 0.6610 V-Cone 30.67 0.6611
" 15.41 0.6610 " 25.75 0.6995
" 15.41 0.242 " 30.69 0.8203
Orice 15.41 0.5000 Wedge 15.41 0.5023
" 20.27 0.6000 " 20.27 0.5023
" 15.41 0.6500 " 20.27 0.6110
" 15.41 0.7000
Table 2
Comparison of ow rates with and without varying discharge coefcients.
Reynolds
number
Flow rate
Venturi w/
constant C (l/s)
Flow rate
Venturi w/
varying C (l/s)
% difference
constant C
to varying C
Flow rate
V-cone w/
constant C (l/s)
Flow rate
V-cone w/
varying C (l/s)
% difference
constant C
to varying C
Flow rate
wedge w/
constant C (l/s)
Flow rate
wedge w/
varying C (l/s)
% difference
constant C
to varying C
100,000 18.84 18.78 0.33% 31.60 31.54 0.19% 20.74 20.85 0.51%
10,000 329.24 325.35 1.20% 661.64 635.12 4.18% 210.86 209.89 0.46%
5000 166.49 162.67 2.35% 334.48 317.56 5.33% 41.56 41.98 1.00%
500 21.22 16.27 30.46% 36.54 31.76 15.06% 20.93 20.99 0.27%
100 4.39 3.25 34.79% 9.53 6.35 50.04% 4.36 4.20 3.94%
10 0.78 0.33 140.07% 2.47 0.64 289.50% 0.76 0.42 81.80%
1 0.21 0.03 557.01% 0.78 0.06 1124.92% 0.23 0.23 448.25%
Table 3
Summary of model discretization uncertainty.
Discretization numerical uncertainty
Venturi Orice V-cone Wedge
Maximum errors
Approximate 0.65% 1.34% 0.83% 1.13%
Extrapolated 0.71% 0.29% 2.33% 2.17%
GCI 1.10% 0.08% 2.15% 0.64%
563 C.L. Hollingshead et al. / Journal of Petroleum Science and Engineering 78 (2011) 559566
Table 4
Venturi meter discharge coefcients.
Venturi
Re Smooth beta=0.661 Sharp beta=0.661 Beta=0.24 Re Smooth beta=0.661 Sharp beta=0.661 Beta=0.24
C C C C C C
50,000,000 0.975 0.967 0.974 500 0.881 0.883 0.879
10,000,000 0.974 0.967 0.974 300 0.841 0.841 0.855
1,000,000 0.973 0.967 0.973 200 0.803 0.801 0.831
100,000 0.971 0.965 0.971 100 0.727 0.715 0.774
75,000 0.970 0.963 0.971 80 0.705 0.688 0.758
50,000 0.967 0.962 0.970 60 0.679 0.650 0.734
30,000 0.967 0.959 0.969 40 0.625 0.596 0.694
10,000 0.961 0.952 0.962 30 0.586 0.554 0.665
5000 0.954 0.944 0.959 20 0.515 0.498 0.621
3000 0.944 0.937 0.943 10 0.432 0.401 0.540
2000 0.937 0.930 0.933 5 0.336 0.300 0.458
1000 0.921 0.914 0.913 1 0.163 0.146 0.254
Table 5
Orice plate discharge coefcients.
Orice
Re Beta=0.50 Beta=0.60 Beta=0.65 Beta=0.70 Re Beta=0.50 Beta=0.60 Beta=0.65 Beta=0.70
C C C C C C C C
50,000,000 0.611 0.595 0.607 0.604 200 0.699 0.725 0.764 0.795
10,000,000 0.613 0.602 0.610 0.607 100 0.702 0.727 0.754 0.781
1,000,000 0.615 0.605 0.612 0.616 80 0.700 0.725 0.748 0.772
100,000 0.619 0.610 0.622 0.623 60 0.697 0.721 0.738 0.756
10,000 0.627 0.624 0.641 0.646 40 0.688 0.707 0.715 0.726
5000 0.632 0.633 0.656 0.659 30 0.677 0.689 0.692 0.696
3000 0.639 0.642 0.666 0.670 20 0.654 0.657 0.648 0.644
2000 0.648 0.658 0.685 0.700 10 0.585 0.568 0.546 0.532
1000 0.670 0.688 0.736 0.765 5 0.478 0.448 0.425 0.407
500 0.684 0.707 0.755 0.788 1 0.233 0.212 0.202 0.191
300 0.693 0.719 0.763 0.796
Table 6
Wedge meter discharge coefcients.
Wedge
Re 20 cm beta=0.5023 15 cm beta=0.5023 Beta=0.611 Re 20 cm beta=0.5023 15 cm beta=0.5023 Beta=0.611
C C C C C C
50,000,000 0.733 0.732 0.705 100 0.699 0.700 0.672
10,000,000 0.732 0.729 0.695 80 0.690 0.692 0.663
1,000,000 0.731 0.729 0.699 60 0.674 0.677 0.645
100,000 0.732 0.730 0.702 40 0.641 0.645 0.606
5000 0.729 0.726 0.700 30 0.610 0.613 0.567
500 0.730 0.725 0.705 20 0.551 0.555 0.503
400 0.725 0.723 0.700 10 0.432 0.433 0.384
300 0.721 0.721 0.694 5 0.318 0.319 0.280
200 0.716 0.716 0.688 1 0.145 0.146 0.127
Table 7
V-cone meter discharge coefcients.
V-cone
Re Beta=0.6611 Beta=0.6995 Beta=0.8203 Re Beta=0.6611 Beta=0.6995 Beta=0.8203
C C C C C C
50,000,000 0.802 0.787 0.734 500 0.700 0.682 0.644
10,000,000 0.805 0.790 0.741 300 0.661 0.645 0.605
1,000,000 0.803 0.792 0.730 200 0.622 0.615 0.557
100,000 0.804 0.792 0.730 150 0.588 0.584 0.526
30,000 0.789 0.781 0.731 100 0.537 0.533 0.472
20,000 0.778 0.774 0.725 80 0.506 0.502 0.440
10,000 0.773 0.766 0.723 60 0.462 0.458 0.400
7500 0.767 0.763 0.724 40 0.396 0.394 0.343
5000 0.765 0.757 0.722 30 0.349 0.350 0.303
4000 0.763 0.755 0.721 20 0.289 0.292 0.253
3000 0.759 0.750 0.714 10 0.207 0.210 0.182
2000 0.750 0.742 0.705 5 0.147 0.150 0.128
1000 0.727 0.721 0.685 1 0.066 0.067 0.057
564 C.L. Hollingshead et al. / Journal of Petroleum Science and Engineering 78 (2011) 559566
for each equation. Fig. 7 shows the correction factor equations plotted
for each meter type for the corresponding equation. If desired, the
reader may use the data in Figs. 36 or in Tables 47 to obtain specic
correction factors.
Venturi meter:
C
r tIn Rev v In Re
2
x In Re
3
1 s In Re u In Re
2
w In Re
3
9
Wedge meter:
C
r tIn Re v In Re
2
1 s In Re u In Re
2
w In Re
3
10
Orice meter:
C
r tIn Re v In Re
2
x In Re
3
z In Re
4
1 s In Re u In Re
2
w In Re
3
y In Re
4
11
V-cone meter:
C
1
r sIn Re =Re t=
Re
p
=Re
:
12
5. Using the results
Figs. 36 demonstrate that the meter geometry, uid viscosity and
measured differential head must be known before calculations can
be completed. The viscosity can be found in published tables or by
using either a viscometer or a rheometer depending on the uid
being measured (Bandyopadhyay and Das, 2007). This information
is needed to obtain an estimation of the average Reynolds number
owing through the ow meter.
The following example problem shows how this research can be
used to obtain a more accurate measure of the ow. For this particular
example a wedge ow meter will be used, but the same process can
be used for any of the ow meters in the study, with only the equa-
tion dening being the primary variation. Eqs. (3) and (4) will be
used to determine the ow rate and the value of the wedge ow
meter.
To show how one could use the data, assume a pipeline with an
inlet diameter (D) of 15.41 cm, which transports heavy oil with a ki-
nematic viscosity of 2.6410
4 m
2
/s. The differential pressure ow
meter being used is a wedge ow meter with an H/D ratio of 0.4. It
is observed that the differential head being produced by the ow
meter is 1.27 cm of water. The process to obtain the ow rate through
the ow meter is as follows:
1. Determine the value, N constant, and d for the particular ow
meter and units in question:
The use of Eq. (2) gives a =0.6111, and Eq. (6) results in d=9.42
2. Guess Re (can get reasonably close by noticing pressure differen-
tial):
Re=1000
3. From relevant gure or equation, for this case Fig. 5, obtain a C
from the discharge coefcient curve for similar values:
Re=1000, C=0.70.
4. Calculating q using Eq. (1), results in a ow of 2.625 l/s.
5. Calculating the average pipe velocity by dividing the ow in centi-
meters by the pipe area results in 0.141 m/s.
6. Computation of the new Reynolds number using average velocity,
inlet diameter, and kinematic viscosity results in a value of 82.3.
7. Use the computed value from step 6 and repeat the iterative pro-
cess steps 3 through 6 until the estimated ow no longer changes
signicantly. For this example, 4 iterations resulted in a ow of
2.513 l/s. The solution of Eq. (1) and the curve from Fig. 5 con-
verged to a ow of 2.475 l/s.
It is signicant to observe that using the correct discharge coef-
cient is vital, as for this example if the constant C=0.70 is used rather
than the calculated C=0.66 it would have resulted in q being 6.1%
greater than the actual ow. The varying discharge coefcient is obvi-
ously an essential parameter for accurate measurement of low Re
ows. The aforementioned process is similar for the other three dif-
ferential pressure ow meters in the study, except the equation
for each meter is different as demonstrated in Eqs. (4)(7).
0.00
0.20
0.40
0.60
0.80
1.00
1.20
1.40
1 10 100 1,000 10,000 100,000 1,000,000 10,000,000 100,000,000
D
i
s
c
h
a
r
g
e
C
o
e
f
f
i
c
i
e
n
t
C
o
r
r
e
c
t
i
o
n
F
a
c
t
o
r
Reynolds Number
Venturi
Orifice
V-cone
Wedge
Fig. 7. Average discharge coefcient correction factors.
Table 8
Coefcients for Eqs. (9)(12).
Coefcient Venturi Wedge Orice V-cone
r 0.19319 0.19865 0.34766 1.00836
s 0.06846 0.28958 0.44896 10.78179
t 0.10127 0.03580 0.03451 11.30940
u 0.00048 0.09657 0.15751
v 0.01168 0.07169 0.08035
w 0.00408 0.00074 0.02469
x 0.00437 0.01895
y 0.00142
z 0.00128
R
2
0.9999 0.9988 0.9994 0.9953
565 C.L. Hollingshead et al. / Journal of Petroleum Science and Engineering 78 (2011) 559566
6. Uncertainty of results
The discretisation errors for each of the models were estimated
using a well-established procedure for reporting uncertainty in CFD
results (Celik et al., 2008). To implement this procedure, three
meshes with a signicantly different number of cells were created
for each of the models. For example one of the V-cone models had a
volume of 0.0519 m
3
and the number of cells in the 3 meshes were;
490,000, 157,000, and 90,000, respectively. Results obtained from
each of the different meshes were used to determine the approximate
error, extrapolated error, and a grid convergence index error. The
largest of the calculated errors was the estimated numerical error
for the model. The study determined that throughout the 13 sets of
computations 42% of the results were within 1%, 83% within 2%, and
95% within 2.5% numerical uncertainty. The results of the grid rene-
ment study are summarized in Table 3. In addition, the physical labo-
ratory data that was used for numerical comparison in the study had
a maximum uncertainty of 0.5%.
7. Conclusions
Discharge coefcients for differential pressure ow meters are
often used in many pipeline transportation industries. This study ex-
amined four different types of ow meters which were tested under
both laminar and turbulent ow conditions. The Venturi, V-cone
and wedge owmeters all have nearly constant discharge coefcients
for moderate to high Reynolds number range applications, but with
the low Reynolds number results presented in this study discharge
coefcients over an expanded range will benet those seeking further
renements in ow measurement. The range of Reynolds numbers
for which physical data was obtained is small in comparison to the
range of data obtained using computational uid dynamics tech-
niques. It is important to understand that this paper characterizes
the performance as a function of Re and does not assume to present
exact C values for each meter type. Physical laboratory testing is re-
quired to determine a precise constant discharge coefcient for indi-
vidual meters. The intent of the study is to present characteristic
curves to enable users to better understand the relative differences
to be expected at low Re.
The accuracy of the CFD results should be greatest at Reynolds
numbers of less than 2300, because turbulence models, with their in-
herent uncertainty, were not implemented in this region. The data
from this research demonstrates that with possible discharge coef-
cients near 0.20 for some of the ow meters, an iterative process
should be used to minimize ow measurement errors. It should be
noted, however, that since only a limited number of ratios were
tested for each of the meters, there may be a variation of results
when using ratios found outside of the study range.
List of symbols
A Constant in Eq. (1)
B Constant in Eq. (1)
Ratio of effective throat diameter to inlet diameter
v-cone
Ratio of effective throat diameter to inlet diameter
C Discharge coefcient for ow meter
C