Flow Measurement and Instrumentation 82 (2021) 102043

Contents lists available at ScienceDirect

Flow Measurement and Instrumentation

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

The calculation of the uncertainty of the orifice-plate discharge coefficient

Michael Reader-Harris *, Claire Forsyth, Tariq Boussouara
National Engineering Laboratory, East Kilbride, Glasgow, G75 0QU, UK

A R T I C L E I N F O A B S T R A C T

Keywords: The uncertainty of the orifice-plate discharge coefficient given by the Reader-Harris/Gallagher (1998) Equation
Orifice plate has been calculated taking account of the uncertainty of the data on which it is based and of the variability in
Differential pressure meter manufacture permitted by ISO 5167–2. This work has shown that using the correct method to determine the
Uncertainty
uncertainty in ISO 5167-2 has made an insignificant difference to the value given in the standard. However, in
other similar cases where the uncertainty for an artefact is based on data from other similar artefacts the un­
certainty values obtained by the correct method may be significantly different from those by the incorrect
method.

1. Introduction to the application 2. Specification of the measurand

It is estimated that at least 40% of industrial flowmeters in use at The measurand is the mass flowrate, as calculated in accordance with
present are differential-pressure-based devices, with the orifice plate ISO 5167–2: 2003 using the discharge coefficient given by the Reader-
being the most popular for accurate measurement of fluid flow. These Harris/Gallagher (1998) Equation (see 3.2).
devices have been described in standards for many years: ISO 5167 [1,2]
now provides the most wide-ranging of these standards. ISO 5167 is used 3. Measurement model
in the measurement of many different fluids across a wide range of ap­
plications; nevertheless, the measurement of natural gas is highest in 3.1. Fundamental equation
terms of financial importance: in 2006 it was estimated that about £16
billion (€20 billion) of natural gas each year in the UK was measured by The attraction of using an orifice meter is that it is very rarely
devices conforming to this standard [3]. necessary to calibrate it in a flowing fluid. It is usual to make it in
A diagram of an orifice meter is provided in Fig. 1. Differential accordance with the pattern in ISO 5167 and then to calculate the mass
pressure can be measured using one of three permissible tapping pairs: flowrate from Equation (1) using measurements made in the factory,
corner tappings (in the corners), D and D/2 tappings (one diameter measurements made in the field and equations in ISO 5167:
upstream and half a diameter downstream of the orifice plate) and
Cε πd2 √̅̅̅̅̅̅̅̅̅̅̅̅̅
flange tappings (one inch (25.4 mm) both upstream and downstream of qm = √̅̅̅̅̅̅̅̅̅̅̅̅̅4 2ρ1 Δp (1)
the orifice). Flange tappings are the most convenient and most common. 1− β 4
Data taken with Corner 2 tappings (a different method of inserting
where qm is the mass flowrate, C is the discharge coefficient (given in ISO
corner tappings designed by Gasunie: see page 131 of [4], where they
5167–2 and discussed in 3.2), ε is the expansibility factor (given in ISO
are called ‘Corner (GU)’) have not been analysed in this report; in every
5167–2), ρ1 is the density at the upstream tapping, Δp is the differential
case where they were included standard corner tappings were present
pressure, and β is the diameter ratio, equal to d/D, where d and D are the
also.
orifice and pipe diameters.
NOTE ‘Orifice plate’ can refer just to the plate or to the whole meter;
where it is important to be clear that the plate and pipework are meant,
‘orifice meter’ can be used.

* Corresponding author.
E-mail addresses: [email protected] (M. Reader-Harris), [email protected] (C. Forsyth), [email protected]
(T. Boussouara).

https://doi.org/10.1016/j.flowmeasinst.2021.102043
Received 3 June 2021; Received in revised form 30 July 2021; Accepted 28 August 2021
Available online 10 September 2021
0955-5986/© 2021 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).
M. Reader-Harris et al. Flow Measurement and Instrumentation 82 (2021) 102043

Fig. 2. Diagram of an orifice plate.

Fig. 1. The approximate pressure profile through an orifice meter.

3.2. Reader-Harris/Gallagher (1998) equation

Using ISO 5167–2: 2003 the discharge coefficient is given by the

Reader-Harris/Gallagher (1998) Equation [5]:

C=0.5961+0.0261β2 − 0.216β8 C∞ term

( 6 )0.7 ( 6 )0.3
10 β 10
+0.000521 +(0.0188+0.0063A)β3.5 slope term
ReD ReD
( ) 4
β
+ 0.043+0.080e− 10L1 − 0.123e− 7L1 (1− 0.11A) upstreamtapping term
1− β4
( ′ ′ 1.1 ) 1.3
− 0.031 M2 − 0.8M2 β downstreamtapping term
(2a)

Where D < 71.12 mm (2.8′′ ) the following term should be added to

Equation (2a):
( )
D
+ 0.011 (0.75 − β) 2.8 − small pipe diameter term (2b) Fig. 3. An EEC orifice plate.
25.4

where D is the pipe diameter in mm. 3.3. Calculating the uncertainty

In equation (2a)
( )0.8 The uncertainty in the discharge coefficient is often the largest
A=
19000β
and M2’ =
2L’2 component of the uncertainty in the flowrate and thus of great impor­
ReD 1− β tance for fair taxation and custody transfer. However, it is not simple to
The Reynolds number, ReD, can be obtained from: calculate the uncertainty for the orifice discharge-coefficient equation.
Evaluating the uncertainty for that equation is different from most
ReD =
4qm evaluations of uncertainty: a normal evaluation of uncertainty is for an
πμ1 D instrument which is calibrated and itself subsequently used. However,
the orifice discharge-coefficient equation is used for plates other than
where μ1 is the dynamic viscosity at the upstream tapping. those used to collect the data.
L1 is the quotient of the distance of the upstream tapping from the Because the standard deviation in the EEC/API database does not
upstream face of the plate necessarily reflect the variability permitted by the standard it is not
and L′ 2 is the quotient of the distance of the downstream tapping appropriate to evaluate the expanded uncertainty, corresponding to
from the downstream face of the plate. 95% coverage, of the orifice discharge-coefficient equation as twice the
Equations (2a) and (2b) are based on the EEC/API database, which is standard deviation of the data in the EEC/API database about the
described in pages 131 to 133 of [4]. These equations represent model equation, which is the method used by API 14.3:1990 [6].
fits to the data, the model being a hybrid based on the physics and some Rather than simply taking twice the standard deviation it is desirable
empirical terms. Where flange tappings are used it is required that both first to reduce as much as possible the effects of random errors in the
ReD ≥ 5000 and ReD ≥ 170β2D, where D is expressed in millimetres. database (see 3.4), then to determine what the uncertainty would be for
To achieve this discharge coefficient and its associated uncertainty an orifice plate of average edge sharpness in a pipe of average rough­
there are many specific requirements in ISO 5167–2 [2] on the di­ ness, taking into account the uncertainty of the original data as well as
mensions and geometry of both the orifice plate (depicted in Fig. 2) and the spread of the data about the equation. However, there is variability
the upstream and downstream pipework. There is a picture of one of the within prescribed limits permitted in the manufacture of orifice meters:
original EEC plates in Fig. 3. so the uncertainty of an orifice meter in accordance with ISO 5167-2
should have additional uncertainty incorporated. In this way the

2
M. Reader-Harris et al. Flow Measurement and Instrumentation 82 (2021) 102043

uncertainty of the Reader-Harris/Gallagher (1998) Equation can be difference between taking the average of the deviations and calculating
calculated. the deviation at the fitted point is less than 0.001%. The uncertainty of
NOTE: It is possible to look at the deviations between orifice plates the fitted line at the fitted point has not been included, since the fitted
that were not used to derive the equation and the equation itself point will not have a larger uncertainty than the individual points do.
(Appendix A), but there is less information about these plates and the
uncertainty of their calibration, and the permitted variability should still 3.5. Errors if data had average pipe roughness and edge sharpness
be included.
The database contains points with a range of values of relative pipe
roughness, Ra/D, and relative edge sharpness, re/d: the fitted equation
3.4. Reducing the effects of random errors in the database
has effectively been derived for a particular relative pipe roughness (as a
function of Reynolds number) and relative edge sharpness (based on the
For each orifice meter (considering an orifice plate with three
relative pipe roughness and relative edge sharpness of the original data);
different sets of tappings as three orifice meters) using a single point
if the actual orifice meter to be manufactured for a user of ISO
from each laboratory means that the points are independent and can be
5167–2:2003 had the same relative pipe roughness and the same rela­
combined as in 3.6. Moreover, this has the advantage described in this
tive edge sharpness as those on which the fitted equation is based, the
section.
uncertainty could be considerably reduced.
Twice the standard deviation of the data in the database about the
The Reader-Harris/Gallagher (1998) Equation is associated with the
Reader-Harris/Gallagher (1998) Equation includes the effects of random
relative pipe roughness (here expressed as k/D) and friction factor λ
errors in the original sets of data collected with equipment whose
shown in Table 9 of ISO/TR 12767:2007 [7], given here as Table 1.
random scatter may be higher than that available today. To avoid the
The relative edge sharpness associated with the Reader-Harris/
uncertainty of the equation being increased by the uncertainty due to
Gallagher (1998) Equation can be calculated from the measurements
random errors in each set of data in the database, it is possible to fit each
of edge sharpness for each plate. The average reduces as the Reynolds
set of data in the database as a linear function of ReD− 0.5 and then
number increases: see Fig. 5, on which the following equation is plotted:
represent each line by a single point at the middle of its range. Such
fitted points have reduced uncertainty due to random errors. Fig. 4 1000
re
= 0.6343e− 0.139lg(ReD )
(3)
(discussed below) illustrates this well. d
Since uncertainties increase for large β, small Red (=ReD/β), small d, Taking the fitted points in the database, calculating (using the cal­
and large Δp/p1 (where p1 is the static pressure at the upstream tapping) culations in 3.7.2 and 3.7.4) what the value would have been with the
the data for 0.19 < β < 0.67, Red > 150000β, d > 50 mm and, if gas, Δp/ relative pipe roughness in Table 1 and the edge sharpness in Equation
p1 ≤ 0.04 were analysed as a core population (NIST Gaithersburg data (3) (where there are multiple points taken in the same laboratory taking
with Δp less than 180 mbar were excluded: see 3.6. Data taken to averages to include all the points) gives the values that are included in
determine installation effects were also excluded). For this core popu­ subsequent sections.
lation of 7031 points the standard deviation, s, about the equation was To the API data no correction was made for edge sharpness.
determined: 2s = 0.41%; there are 339 data sets making up the core
population: for the 339 mid-range calculated points 2s = 0.38%. To 3.6. Calculating the deviations taking account of the uncertainty of the
illustrate the difference in deviations Fig. 4 shows one set of 24 water laboratories
points from the database. Although twice the r.m.s. (root mean square)
deviation of the (24) points about the Reader-Harris/Gallagher (1998) It is necessary also to consider the uncertainty contributed by the
Equation is 0.28%, the (expanded) random uncertainty of the fitted original laboratories: on the one hand the uncertainty of the average of
point is 0.06%. The deviation of the fitted point from the Reader-Harris/ the data from the independent laboratories should be added to the un­
Gallagher (1998) Equation and the mean deviation of the 24 points from certainty of the Equation; on the other hand the scatter contributed by
the Reader-Harris/Gallagher (1998) Equation are both equal to − calibration in a range of independent laboratories should be removed
0.03%. from the uncertainty of the Equation.
In Fig. 4 the fitted straight line is almost exactly parallel to the Almost all the European orifice plates were calibrated in multiple
Reader-Harris/Gallagher (1998) Equation; i.e. the error varies very little laboratories. The errors for the orifice plates in 24” (600 mm) pipe are
with Reynolds number. At the fitted point the fitted discharge coefficient shown in Fig. 6. Data were collected by Delft Hydraulics, NEL (water),
is equal to the average value of the measured discharge coefficients. The British Gas and Gasunie.
In Fig. 6 for β = 0.37, 0.5 and 0.57 the deviations are consistent and
have little variation with Reynolds number. It is less obvious whether
the same is true for β = 0.2 and 0.66. Testing the data with the χ 2 test in
5.3 of Cox [8] gives χ 2obs < 1.01 for β = 0.37, 0.5 and 0.57, χ 2obs < 3.34
for β = 0.66, and 8.8 < χ 2obs < 11.4 for β = 0.2. The consistency check in
Cox is met for all β except 0.2. Removing either the highest set or the
lowest one for β = 0.2 gives consistency: removing the lowest one gave
slightly lower χ 2obs.
A similar test is applied to the EEC 10” (250 mm) data: see Fig. 7.
Data were collected by Delft Hydraulics, NEL (water), British Gas, CEAT,
Gasunie and Ruhrgas. The results were found to be consistent for all
pairs of tappings (except for a plate that was subsequently replaced.
Only the data from the replacement plate are included in Fig. 7 or in
subsequent calculations). ‘0.57*’ refers to a second orifice plate with β
= 0.57.
The fact that the data are consistent in the 10” (250 mm) and 24”
(600 mm) sets suggests that the claimed uncertainties are not too small
Fig. 4. One set of National Engineering Laboratory (NEL) 4” (100 mm) β = (at least after some of the random uncertainty has been removed).
0.57 water data from the orifice-plate discharge-coefficient database. A similar test is applied to the EEC 4” (100 mm) data: see Fig. 8. Data

3
M. Reader-Harris et al. Flow Measurement and Instrumentation 82 (2021) 102043

Table 1
Values of k/D and λ associated with the Reader-Harris/Gallagher (1998) Equation.
ReD 104 3 × 104 105 3 × 105 106 3 × 106 107 3 × 107 108
4
10 k/D 1.75 1.45 1.15 0.9 0.7 0.55 0.45 0.35 0.25
λ 0.031 0.024 0.0185 0.0155 0.013 0.0115 0.0105 0.010 0.0095

Fig. 5. Edge sharpness (re) relative to the orifice diameter for each set of EEC
data: values are plotted at the pipe Reynolds number of the fitted point for
the set. Fig. 7. Discharge coefficient deviation from the Reader-Harris/Gallagher
(1998) Equation for the EEC 10” (250 mm) data adjusted to average pipe
roughness and edge sharpness.

Fig. 6. Discharge coefficient deviation from the Reader-Harris/Gallagher

(1998) Equation for the EEC 24” (600 mm) data adjusted to average pipe
roughness and edge sharpness (in this key and those for Figs. 7 and 8 the value
Fig. 8. Discharge coefficient deviation from the Reader-Harris/Gallagher
of β and the pressure tappings used are given).
(1998) Equation for the EEC 4” (100 mm) data adjusted to average pipe
roughness and edge sharpness.
were collected by NEL (water), Gasunie, Gaz de France and NEL (air).
The results are only found to be consistent for all three tappings once the from NIST Gaithersburg.
second set of three points from the right was removed for both β = 0.57 The weighted average deviations for the consistent sets are shown in
and β = 0.66 and the left hand set of three points removed for β = 0.57. Fig. 9. Each point represents a different orifice meter (considering an
The NEL water data and NEL air data are almost entirely independent of orifice plate with three different sets of tappings as three orifice meters).
each other (and, after removal of points as above, are only both used for The weighted averages were obtained using the method in 5.1 of Cox
β = 0.66). [8].
The 4” (100 mm) data that were collected to study installation effects In support of the idea that a plate has a single value of deviation some
have been omitted. modern NEL data are provided in Appendix B: the data are roughly
The NIST Gaithersburg data have an uncertainty that increases very parallel to the Reader-Harris/Gallagher (1998) Equation. Moreover, in
rapidly at low differential pressure. By limiting the data to those with Figs. 6–8, roughly the same number of deviations have a positive slope
differential pressure above 180 mbar each of the averaged data has an against Reynolds number as have a negative slope. Even if it were not
uncertainty within the range 0.125%–0.145%. accepted that the deviation can be considered constant then for each
If three points or fewer for a single plate from a laboratory met the plate the weighted mean deviation would apply for an intermediate
criteria in 3.4 the averaged point was not used; so all the API data were value of Reynolds number.

4
M. Reader-Harris et al. Flow Measurement and Instrumentation 82 (2021) 102043

Fig. 10.
Fig. 10 gives measured and computed (using CFD) values of ΔC as a
function of β3.5Δλ, where λ is the friction factor given by

λρu2 x
Δp = (4)
2D

where Δp is the difference in pressure between two tappings spaced a

distance x apart in a pipe of diameter D and u is the mean velocity in the
pipe.
References to the computational and experimental data are given in
Reader-Harris [4], which also includes the derivation of Equation (5):

(5)
′
ΔCrough = (3.134 + 4.726A )β3.5 Δλ

where ΔCrough is the change in discharge coefficient due to a change in

friction factor Δλ, and A′ is given by
( )0.9
Fig. 9. Weighted average deviations for the consistent sets. ′
A =
2100β
. (6)
ReD
P, the maximum permissible percentage shift in C, is given in Table 3.
Table 2
Deviations of the fitted and adjusted points from the Reader-Harris/Gallagher U, the stated percentage uncertainty of the Reader-Harris/Gallagher
(1998) Equation. (1998) Equation in ISO 5167–2, is also given in Table 3, for P was
determined to ensure that for β ≤ 0.5, where other sources of error are
β < 0.6 β ≈ 0.66
dominant, P/U ≤ β; for 0.5 < β ≤ 0.71 P/U = 0.5; for 0.71 < β P/U
Mean deviation (%) 0.001 − 0.004 increases from 0.5 at β = 0.71 to 0.55 at β = 0.75.
2 × r.m.s. deviation (%) 0.220 0.288
NOTE It can be checked that this value of P is correct, since it,
Expanded uncertainty of points (%) 0.141 0.130
Square root of the sums of the squares of the two lines above 0.261 0.315 together with the value of λ associated with the data in the database (and
(%) thus with the Reader-Harris/Gallagher (1998) Equation) in Table 1, and
Equation (5) to calculate the effect of Δλ on C, leads to the maximum and
minimum values of pipe roughness for pipes upstream of orifice plates
The deviations of the points in Fig. 9 are summarized in Table 2. given in Tables 1 and 2 of ISO 5167–2:2003 (except that the maximum
The expanded uncertainty of the points was obtained by evaluating value of Ra/D has been set as 1.5 × 10− 3; so in some cases the maximum
the uncertainty for each point in Fig. 9 using the method in 5.2 of Cox deviation is smaller than P above).
[8] and then taking the r.m.s. value of all these uncertainties. The
expanded uncertainty of the individual points in Fig. 9 varied from 3.7.3. Installation effects
0.11% to 0.16%. The requirement for upstream lengths used by ISO 5167–2:2003 was
The API data (with no averaging between facilities) give roughly the put forward by API: all the installation-effects data of sufficient quality
same deviations as the EEC data when all the fitted and adjusted data are brought together; at least two high-quality sets of data are necessary
points in each set are considered. to determine each upstream length; the permissible upstream length
chosen for the standard is a location at which data were taken; more­
3.7. Variability permitted by ISO 5167 over, the data taken at that location and all the data with longer up­
stream lengths must lie within the acceptable band; the acceptable band
3.7.1. Introduction is half the (expanded) uncertainty of the Reader-Harris/Gallagher (RG)
What is required as the uncertainty of the discharge-coefficient Equation at infinite Reynolds number as calculated by API (from the
equation is the uncertainty of an orifice meter with the variability scatter of the original data about the Equation). The value of this
permitted by the standard. (expanded) uncertainty is shown in Figs. 1–4 of API MPMS 14.3.1:1990
(Fig. 4 of API MPMS 14.3.1:2012 [9]): it is.
3.7.2. Pipe roughness
The effect of pipe roughness on discharge coefficient is given in • 0.51% at β = 0.2
• 0.44% at β = 0.5
• 0.44% at β = 0.6
• 0.47% at β = 0.67
• 0.56% at β = 0.75.

NOTE By using this rule the maximum shift due to upstream instal­
lation should in practice often be significantly less than the maximum

Table 3
Maximum permissible percentage shift, P, in C due to pipe roughness.
β P U

0.1 ≤ β < 0.2 0.5β 0.7 - β

0.2 ≤ β ≤ 0.5 0.5β 0.5
0.5 < β ≤ 0.6 0.25 0.5
0.6 < β ≤ 0.71 0.5 (1.667β - 0.5) 1.667β - 0.5
0.71 < β 1.13β3.5 1.667β - 0.5
Fig. 10. The effect of rough pipe on the orifice plate discharge coefficient.

5
M. Reader-Harris et al. Flow Measurement and Instrumentation 82 (2021) 102043

permitted value (i.e. half of the (expanded) uncertainty shown in

Figs. 1–4 of API MPMS 14.3.1:1990).

3.7.4. Edge sharpness

The permissible value of edge sharpness is given by re ≤ 0.0004d. The
effect of edge radius is given in Fig. 11, which shows the data of Hobbs
and Humphreys [10], taken in 12” (300 mm) pipe. The data points with
the smallest value of re/d for each β are taken as having no shift in
discharge coefficient. The slope of the fitted line is approximately 550;
so the effect of a change of edge radius of 0.00027d on C is 0.15%.
A maximum effect of edge sharpness of 0.15% gives the dashed-dot
line shown in Fig. 5. Moreover, all the points in the database are be­
tween the dashed line and the maximum permissible value. For ReD
greater than 106 the edge sharpness cannot cause a deviation of more
than 0.15% from the Reader-Harris/Gallagher (1998) Equation. No
actual EEC plates gave a deviation due to edge sharpness more than Fig. 12. Upstream steps: % shift in discharge coefficient due to pipes at various
0.09% from the Reader-Harris/Gallagher (1998) Equation. distances upstream of an orifice plate (β = 0.67) in a Schedule 40 pipe.

3.7.5. Plate bending pipe and Schedule 80 pipe, respectively.

With a slope of 1%, the theoretical model of Jepson and Chipchase
[11,12] gives a shift in discharge coefficient of a little less than 0.2% in 4. Uncertainty propagation
absolute value at maximum. In practice an orifice meter system is
checked so that it is known that the plate has a slope of less than 0.5% (i. All the uncertainty components in 3.6 and 3.7 are listed in Table 4.
e. a shift in discharge coefficient of less than 0.1%) when there is no If the components in Table 4 (with 0.15% for edge sharpness) are
differential pressure across the plate (ISO 5167–2:2003 5.1.3.1). The combined by taking the root sum square of the components in Table 4
plate thickness will have been chosen so that elastic deformation will not the total expanded uncertainty increases monotonically from 0.44 (0) %
give a shift in discharge coefficient of more than 0.1% (i.e. the plate has to 0.48 (4) % as β increases from 0.2 to 0.6 and then to 0.55 (8) % at β =
a slope of less than 0.5% due to elastic bending: see ISO/TR 9464:2008 0.67. All these are a little smaller than the uncertainty values in ISO
5.2.5.1.2.3 [13]). 5167–2:2003 (see Table 3).
It is assumed (reasonably) that the two types of bending, i.e. at If, following the GUM [15], the standard uncertainty is obtained
manufacture and due to differential pressure, are independent of each from the expanded uncertainty by dividing by 2 for the base uncertainty
other. (since the probability distribution is assumed to be normal) and by √3
for the other components (since their probability distributions are
3.7.6. Eccentricity assumed to be rectangular) the components are given in Table 5.
The ISO limit according to Fig. 6 of ISO/TR 12767:2007 [7] gives If the components in Table 5 (with 0.087% for edge sharpness) are
shifts up to about 0.09% in magnitude. combined by taking the root sum square of the components in Table 5
the standard uncertainty increases monotonically from 0.243% to
3.7.7. Steps in pipe size 0.269% as β increases from 0.2 to 0.6 and then to 0.309% at β = 0.67.
Permissible steps between upstream sections of pipe are given in When these are multiplied by 1.96 the expanded uncertainty increases
6.4.3 of ISO 5167–2:2003. The effect of a permissible step is about 0.1% monotonically from 0.476% to 0.527% as β increases from 0.2 to 0.6 and
at worst for β = 0.67: in the case depicted in Fig. 12 (see Reader-Harris then to 0.606% at β = 0.67. Four out of six values are smaller than the
and Brunton [14]) a Schedule 10 pipe is permitted beyond 10D from the uncertainty values in ISO 5167–2:2003 (see Table 3); for β ≤ 0.6 the
orifice plate and a Schedule 80 pipe beyond 28D from the orifice plate average value is within 0.002% of the uncertainty value in ISO
(no length is specified for a Schedule 120 pipe). The effect is about 0.1 5167–2:2003. In many cases there will be no steps; often the plate
(0.6/0.67)3.5 % at worst for β = 0.6, as a dependence on β raised to a thickness will be chosen to avoid plate bending near the maximum
power of about 3–4 is typical for velocity profile effects on orifice plates permitted; in many cases (e.g. in high-pressure natural gas in larger
([4] p. 249) and the minimum lengths are 10D and 26D for Schedule 10 pipes for larger β, where in most cases ReD > 106) the uncertainty due to

Table 4
Calculated percentage expanded uncertainty components for the Reader-Harris/
Gallagher (1998) Equation.
β 0.2 0.3 0.4 0.5 0.6 0.67

Base uncertainty for 0.261 0.261 0.261 0.261 0.261 0.315

calculated pipe
roughness and edge
sharpness
Pipe roughness 0.10 0.15 0.20 0.25 0.25 0.31
Installation effects 0.255 0.24 0.23 0.22 0.22 0.235
Edge sharpnessa 0.15 0.15 0.15 0.15 0.15 0.15
Plate bending at 0.1 0.1 0.1 0.1 0.1 0.1
manufacture
Plate bending due to 0.1 0.1 0.1 0.1 0.1 0.1
differential pressure
Eccentricity 0.09 0.09 0.09 0.09 0.09 0.09
Steps in pipe size 0.003 0.007 0.018 0.039 0.068 0.1
Total 0.440 0.446 0.460 0.481 0.484 0.558
a
Fig. 11. Effect of orifice edge radius (re) on discharge coefficient: NEL tests. 0.15% includes likely practical range; 0.09% includes all the EEC data.

6
M. Reader-Harris et al. Flow Measurement and Instrumentation 82 (2021) 102043

Table 5
Calculated percentage standard uncertainty components for the Reader-Harris/Gallagher (1998) Equation.
β 0.2 0.3 0.4 0.5 0.6 0.67

Distribution

Base uncertainty for calculated pipe roughness and edge sharpness Normal 0.131 0.131 0.131 0.131 0.131 0.158
Pipe roughness Rectangular 0.058 0.087 0.115 0.144 0.144 0.179
Installation effects Rectangular 0.147 0.139 0.133 0.127 0.127 0.136
Edge sharpnessa Rectangular 0.087 0.087 0.087 0.087 0.087 0.087
Plate bending at manufacture Rectangular 0.058 0.058 0.058 0.058 0.058 0.058
Plate bending due to differential pressure Rectangular 0.058 0.058 0.058 0.058 0.058 0.058
Eccentricity Rectangular 0.052 0.052 0.052 0.052 0.052 0.052
Steps in pipe size Rectangular 0.002 0.004 0.010 0.023 0.039 0.058
Total 0.243 0.246 0.255 0.267 0.269 0.309
Expanded uncertainty: total 0.476 0.483 0.500 0.524 0.527 0.606
a
0.087% includes likely practical range; 0.052% includes all the EEC data.

edge sharpness is lower; moreover, in many cases the uncertainty due to Validation, Formal analysis, Writing – Original Draft; Claire Forsyth:
installation effects has been overstated (see 3.7.3). Software; Tariq Boussouara: Software.

5. Reporting the result Declaration of competing interest

The uncertainty of the Reader-Harris/Gallagher (1998) Equation has The authors declare that they have no known competing financial
been calculated taking account of all the sources of uncertainty, interests or personal relationships that could have appeared to influence
including the uncertainty of the original data and the variability the work reported in this paper.
permitted by ISO 5167–2. The value (see Tables 4 and 5) is very similar
to that in ISO 5167–2 (see Table 3); the standard will not be changed. Acknowledgements
Confidence in such an important standard has been increased.
This project has received funding from the EMPIR programme co-
6. Interpretation of results financed by the Participating States and from the European Union’s
Horizon 2020 research and innovation programme.
Although no change to ISO 5167 will be proposed, a standard has The authors are grateful to Maurice Cox (NPL, UK), Álvaro Silva
been set here for other similar calculations where results may be Ribeiro (LNEC, Portugal), João Alves e Sousa (IPQ, Portugal) and
significantly different from those by the incorrect method. Adriaan van der Veen (VSL, Netherlands), who provided valuable
Although the uncertainty of calibration facilities has improved since comments on this work and made it clearer to the reader.
the 1980s it would not be easy to achieve a very large reduction in The EEC data included in this analysis were collected by British Gas,
uncertainty by collecting new data and fitting an improved discharge- CEAT, Delft Hydraulics, Gasunie, Gaz de France, NEL, and Ruhrgas; the
coefficient equation: if the base uncertainty in Table 4 were halved data collection was supported by DG XII of the Commission of the Eu­
the total expanded uncertainty would only reduce by about 0.06% ropean Communities; the original analysis was supported by both DG XII
(unless, for example, the minimum upstream straight lengths were and the National Measurement System Policy Unit of the UK Department
increased or a narrower range of permitted pipe roughness achieved). of Trade and Industry. The American Petroleum Institute (API)-spon­
sored data included in this analysis were collected by NIST
Credit Gaithersburg.

Michael Reader-Harris: Conceptualization, Methodology,

APPENDIX A

Comparison with other data.

One way to determine the uncertainty of the Reader-Harris/Gallagher (1998) Equation is to examine how it performs when the discharge co­
efficients of orifice plates not used to determine the equation are compared with it. Just over 100 runs of calibration data using more than 50 orifice
plates, collected by NEL, NOVA, NIST Boulder, Daniel and SwRI [16–25], have been compared with it: for each set of data a mean deviation has been
calculated. These sets of data include both water data and high-pressure gas data; so there is a useful range of Reynolds number. The mean deviations
are shown in Fig. A.1. From this figure the quoted uncertainty in ISO 5167–2:2003 appears reasonable, but no account is taken of the uncertainty of the
data itself or any analysis of the range of permissible manufactured pipes or plates or installations.

7
M. Reader-Harris et al. Flow Measurement and Instrumentation 82 (2021) 102043

Fig. A.1. Percentage mean deviation in C from the Reader-Harris/Gallagher (1998) Equation from more than 50 orifice plates.

APPENDIX B

Calibrating 8” (200 mm) orifice plates.

Six orifice plates were calibrated in 8” (200 mm) pipe: one β = 0.2, one β = 0.4 and four β = 0.6. The plates had spark-eroded orifices: although no
data on edge sharpness or pipe roughness were collected it has been assumed that the edge radius was 9 μm, the same as on smaller orifices from the
same manufacturer (see p. 71 of [4]) and that the pipes had the roughness given on the drawings (upstream pipes were 0.8 μm or 1.6 μm; 1.2 μm has
been used for the graphs below).
Figures B1 to B.3 show the data for the three values of β: the Reader-Harris/Gallagher (1998) Equation (here called (RG (1998)) is shown on each
together with its uncertainty bands; an adjusted equation taking account of the edge sharpness of the orifice and the roughness of the pipe is also
shown together with an uncertainty of 0.261%. The uncertainty of the data is 0.16%. All the sets of data are close to parallel to the Reader-Harris/
Gallagher (1998) Equation.

Fig. B.1. Data for β = 0.2

8
M. Reader-Harris et al. Flow Measurement and Instrumentation 82 (2021) 102043

Fig. B.2. Data for β = 0.4

Fig. B.3. Data for β = 0.6 ‘Base 1’, ‘Base 7’, ‘Base 9’ and ‘Base thin’ are four different orifice plates. ‘np’ refers to a set of data with different downstream pipework.
‘horiz’ and ‘vert’ refer to horizontal and vertical tappings.

References [7] INTERNATIONAL ORGANIZATION FOR STANDARDIZATION, Measurement of

Fluid Flow by Means of Pressure Differential Devices - Guidelines on the Effect of
Departure from the Specifications and Operating Conditions Given in ISO 5167,
[1] INTERNATIONAL ORGANIZATION FOR STANDARDIZATION. Measurement of
ISO/TR 12767:2007, Geneva, 2007.
Fluid Flow by Means of Pressure Differential Devices Inserted in Circular Cross-
[8] M.G. Cox, The evaluation of key comparison data, Metrologia 39 (2002) 589–595.
Section Conduits Running Full - Part 1: General. Geneva: ISO 5167-1:2003.
[9] AMERICAN PETROLEUM INSTITUTE, Orifice Metering of Natural Gas and Other
[2] INTERNATIONAL ORGANIZATION FOR STANDARDIZATION. Measurement of
Related Hydrocarbon Fluids – Concentric, Square-Edged Orifice Meters, Part 1 -
Fluid Flow by Means of Pressure Differential Devices Inserted in Circular Cross-
General Equations and Uncertainty Guidelines vol. 1, 2012. Manual of Petroleum
Section Conduits Running Full - Part 2: Orifice Plates. Geneva: ISO 5167-2:2003.
Measurement Standards, (Chapter 14).3.1. September 2012: Errata 2013.
[3] M.J. Reader-Harris, ISO flow measurement standards — report on the ISO/TC 30
Washington DC: API 14.3.
meeting in November 2006, Flow Meas. Instrum. 18 (3–4) (2007) 114–120.
[10] J.M. Hobbs, J.S. Humphreys, The effect of orifice plate geometry upon discharge
[4] M.J. Reader-Harris, Orifice Plates and Venturi Tubes, Springer, 2015.
coefficient, Flow Meas. Instrum. 1 (3) (1990) 133–140.
[5] M.J. Reader-Harris, J.A. Sattary, The orifice plate discharge coefficient equation -
[11] P. Jepson, R. Chipchase, The effect of plate buckling on orifice meter accuracy.
the equation for ISO 5167-1, in: Proc. 14th North Sea Flow Measurement
Report No ERS.R.467, British Gas Engineering Research Station, 1973.
Workshop, Peebles: Paper 24, National Engineering Laboratory, East Kilbride,
[12] P. Jepson, R. Chipchase, Effect of plate buckling on orifice meter accuracy, J. Mech.
Glasgow, 1996.
Eng. Sci. 17 (6) (1975) 330–337.
[6] AMERICAN PETROLEUM INSTITUTE, Manual of Petroleum measurement
[13] INTERNATIONAL ORGANIZATION FOR STANDARDIZATION, Guidelines for the
standards, chapter 14 - natural gas fluids measurement, section 3 - concentric,
Use of ISO 5167:2003, ISO/TR 9464:2008, Geneva, 2008.
square-edged orifice meters, Part 1 - general equations and uncertainty guidelines
[14] M.J. Reader-Harris, W.C. Brunton, The effect of diameter steps in upstream
1 (1990). MPMS (Chapter 14), Section 3, Part 1, 3rd Edition. Washington DC: API
pipework on orifice plate discharge coefficients, in: Proc. 5th Int. Symposium Fluid
14.3.
Flow Measurement, Washington DC, 2002.

9
M. Reader-Harris et al. Flow Measurement and Instrumentation 82 (2021) 102043

[15] BIPM, IEC, IFCC, ILAC, ISO, IUPAC, IUPAP, OIML, Guide to the Expression of [21] M.J. Reader-Harris, W.C. Brunton, I.G. Nicholson, R. Rushworth, Ageing effects on
Uncertainty in Measurement, JCGM 100:2008, GUM 1995 with Minor Corrections, orifice metering, in: Proc. 21st North Sea Flow Measurement Workshop, Norway,
BIPM, 2008. 2003.
[16] D.L. George, T.B. Morrow, Orifice meter calibration for backwards-facing orifice [22] M.J. Reader-Harris, D. Hodges, R. Rushworth, The effect of drain holes in orifice
plates. GRI Report No 01/0074 on SwRI Project No 18-8890, Gas Research plates on the discharge coefficient, in: Proc. 26th Int. North Sea Flow Measurement
Institute, Chicago, 2001. Workshop, St Andrews. NEL, East Kilbride, Glasgow, 2008.
[17] T.B. Morrow, Orifice meter expansion factor tests in 4-inch and 6-inch meter tubes. [23] M.J. Reader-Harris, N. Barton, D. Hodges, The effect of contaminated orifice plates
GRI Report GRI-04/0042, Gas Research Institute, Chicago, 2004. on the discharge coefficient, in: Proc. 15th FLOMEKO, Taipei, 2010.
[18] T.B. Morrow, Metering research facility program: additional studies of orifice [24] W. Studzinski, U. Karnik, P. LaNasa, T. Morrow, D. Goodson, Z. Husain,
meter installation effects and expansion factor. GRI Report GRI-04/0246 on SwRI J. Gallagher, White paper on orifice meter installation configurations with and
Project No 18.10315, Gas Research Institute, Chicago, 2005. without flow conditioners, in: White Paper Prepared for API 14.3 Part 2. GRI
[19] T.B. Morrow, J.T. Park, Baseline conditions for orifice meter calibration. GRI Report GRI 99/0262, Gas Research Institute, Chicago, 2001.
Report GRI-92/0097, Gas Research Institute, Chicago, 1992 (as amended by Errata, [25] W. Studzinski, M. Weiss, J. Geerligs, J. Attia, Effect of reducers, expanders, a gate
1993). valve and two bends in perpendicular planes on orifice meter performance, in:
[20] M.J. Reader-Harris, The Effect on Orifice Plates of Diameter Steps in Upstream Proc. Flow Measurement 2001 Int. Conf., Peebles: Paper 3.1, National Engineering
Pipework and of Swirl, East Kilbride, Glasgow, 2002. Report No 2002/75 on Laboratory, East Kilbride, Glasgow, 2001.
Project No FDDP02. NEL.

10