INTERNATIONAL JOURNAL OF PRECISION ENGINEERING AND MANUFACTURING Vol. 18, No. 8, pp.

1075-1083 AUGUST 2017 / 1075

REGULAR PAPER DOI: 10.1007/s12541-017-0126-0

ISSN 2234-7593 (Print) / 2005-4602 (Online)

Water Flow Meter Calibration with a Master Meter Method

Sejong Chun1, Byung-Ro Yoon1, Hae-Man Choi1, and Yong Bong Lee1,#
1 Division of Physical Metrology, Korea Research Institute of Standards and Science, 267, Gajeong-ro Yuseong-gu, Daejeon, 34113, South Korea
# Corresponding Author / E-mail: [email protected], TEL: +82-42-868-5305, FAX: +82-42-868-5643

KEYWORDS: Calibration, Flow meter, Master meter, Measurement uncertainty, Water flow

Master meter method is an efficient way to calibrate flow meters with less time and costs. The master meter method has simpler flow
configuration than the gravimetric flow measurement method requires. Maintenance efforts of the master meter method are easier
than the gravimetric method. However, the guidelines of master meter method have not been established completely. It is because the
notion of relative deviations, which are used as a measure and for comparison calibration, is not clearly understood. On the other
hand, the system uncertainty, which is typically used as the calibration and measurement capability of the gravimetric flow standards,
are understood very well. In this study, a complete explanation on the master meter method is attempted. First, appropriate calibration
conditions are assumed. Second, a suitable calibration procedure is suggested. Third, a mathematical model for the relative deviations
is described. Fourth, uncertainty factors of the relative deviations are evaluated. Finally, two examples of the comparison calibration
are explained in detail. The master meter method is compatible with the guide to the expression of uncertainty in measurement (GUM,
ISO/IEC Guide 98-3:2008). In the examples, it is found that the master meter method can achieve the measurement uncertainty as
low as 0.20%.

Manuscript received: October 11, 2016 / Revised: February 2, 2017 / Accepted: April 17, 2017

NOMENCLATURE m· DUT = mass flow rate of DUT (kg/s)

m· REF = mass flow rate of REF (kg/s)
EA = repeatable measurement of E
EB = correction value of E n = number of measurements
E = relative deviation (%) qDUT = volume flow rate of DUT (m3/h)
Ei = i-th measurement value of E (%) qREF = volume flow rate of REF (m3/h)
E = averaged value of Ei (%) qmax = maximum volume flow rate (m3/h)
K = K factor (pulse/m3) qmin = minimum volume flow rate (m3/h)
U(E) = measurement uncertainty of E (%) r = correlation coefficient
Us = measurement uncertainty of REF (%) tREF = measurement time of qREF (s)
cfREF = sensitivity coefficient of fREF (m3/h/pulse) xi = measurand
u(E) = combined uncertainty of E (%)
ciREF = sensitivity coefficient of iREF (m3/h/mA) u(EA) = type A evaluation of u(E) (%)
cxi = sensitivity coefficient, ∂E / ∂xi u(EB) = type B evaluation of u(E) (%)
ur(qDUT) = relative uncertainty of qDUT (%)
cr xi = relative sensitivity coefficient, ∂E / ∂xi ur(qREF) = relative uncertainty of qREF (%)
fREF = pulse counts of qREF (pulse) u(xi) = standard uncertainty of xi
iREF = current reading of qREF (mA) ur(xi) = relative standard uncertainty of xi
imax = maximum current (mA) ur(δfREF 1) = uncertainty due to calibration certificate (%)
imin = minimum current (mA) ur(δfREF 2) = uncertainty due to long-term deviation (%)
k = coverage factor ur(δiREF 1) = uncertainty due to calibration certificate (%)

© KSPE and Springer 2017

 1076 / AUGUST 2017 INTERNATIONAL JOURNAL OF PRECISION ENGINEERING AND MANUFACTURING Vol. 18, No. 8

such as the long-term stability and the uncertainty of associated

ur(δiREF 2) = uncertainty due to long-term deviation (%) instruments must be included as indicated by the WGFF guideline. In 17

ur(δqDUT) = type B evaluation of ru(qDUT) (%) addition, a cross-correlation term between the REF and the DUT must
ur(δqREF) = type B evaluation of ru(qREF) (%) be included. It is because the flow rates of the REF and the DUT are
ur(δqDUT 1) = uncertainty due to minimum resolution (%) not independent due to the continuity equation connecting two flow
ur(δqREF 1) = uncertainty due to minimum resolution (%) 18
meters along a pipeline. Detailed description on the comparison
ur(δqREF 2) = uncertainty due to calibration certificate (%) calibration is needed to set up the guideline for the master meter
ur(δqREF 3) = uncertainty due to long-term deviation (%) method in order to provide calibration service for custody transfer of
δE = long-term deviation of E (%) natural gas or hydrocarbons flows, and for better regulations on water
δqDUT = correction value of qDUT (m3/h)
10-16
resource management of potable water.
δqREF = correction value of qREF (m3/h) This study focuses on how the comparison calibration can be
described according to “the guide to the expression of uncertainty in
δqmin = resolution of associated instrument (m3/h)
measurement (GUM)” and its related standard documents such as EA-
ρDUT = liquid density of DUT (kg/m3) 7,8
4/02. Toward this end, the measurement uncertainty of the DUT is
ρREF = liquid density of REF (kg/m3) evaluated systematically. First, an appropriate calibration condition is
v = effective degrees of freedom discussed. Second, a calibration procedure is suggested. Third, a
vE = degrees of freedom of E mathematical model for relative deviations is described. Fourth, the
vEA = degrees of freedom of EA uncertainty factors are evaluated by classifying into type A and type B
evaluations of uncertainties. Finally, two examples on evaluating the
vqDUT = degrees of freedom of qDUT
measurement uncertainty of flow meters are explained.
vq = degrees of freedom of qREF
REF

2. Comparison Calibration

2.1 Assumptions for calibration condition

1. Introduction The comparison calibration is to compare two output signals from
a REF and a DUT as illustrated in Fig. 1. The output signals can be
There are many accredited laboratories which choose master meter pulses, which are used to indicate fluid volume in units of m , or 3

method as their working standards for flow metering in the South electric currents, which are used to indicate flow rate in units of m /h. 3

1
Korea. The master meter method has advantages to the gravimetric Since the REF and the DUT is connected along a pipeline, the two
flow measurement standard in various ways, although the gravimetric output signals can be cross-correlated by continuity equation. This 18

flow measurement standard is implemented by many national metrology can raise subtle problems in evaluating the measurement uncertainty
2,3
institutes. First, the master meter method is not very complicated because the two signals are dependent. The relative deviations between
4-6
compared with gravimetric flow standards. It is because a reference the two output signals can avoid such subtle problems.
flow meter (REF) can indicate the reference flow rate which is traceable The continuity equation constrains that mass passing through a
7,8
to the SI units. Second, maintenance costs of the master meter method pipeline must be conserved. This means that volume flow rate can be
are lower than the costs for the gravimetric flow standards. Third, time regarded as being constant as long as density is assumed to be constant
and efforts for calibrating a test flow meter (DUT) can be saved by in the control volume in Fig. 1. Thus, the relative deviations can be
employing the master meter method. described by the volume flow rate in comparison calibration.
The master meter method can be used to calibrate various types of In case of water flows, water density formula can be used to
9-16
flow meters for natural gas, hydrocarbon and potable water. calculate water density by temperature measurement. 19
The
International standard documents on custody transfer require that the temperature range for the water density formula is constrained between
0°C and 40°C. If water temperature is measured within (20 ± 0.2)°C,
11-16
flow meters with class 1 must be accurate as low as ± 0.2%. This 19

means that the measurement uncertainty of the flow meters should be 3

water density is changed from 998.16 kg/m to 998.24 kg/m 3.19
This
less than 0.2% (k = 2) to meet the accuracy requirement. This indicates that the water density is changed less than ± 0.04 kg/m if the
3

uncertainty level can be achieved by the master meter method. water temperature is changed within ± 0.2°C. Likewise, hydrocarbon
The master meter method can be implemented by calibrating the 3
density can be changed between 815.01 kg/m and 815.15 kg/m if 3

flow rate by DUT with the reference flow rate by REF, which is known hydrocarbon temperature is constrained in (20 ± 0.1)°C. This means
20

as the comparison calibration. However, general guidelines for the that the accuracy of hydrocarbon density is less than ± 0.07 kg/m if the
3

accuracy of temperature measurement is within ± 0.1°C. Pressure

17
comparison calibration have not been established completely. For
example, the system uncertainty is well understood in many national conditions can be also interested with this manner. Pressure range of
3
metrology institutes. The system uncertainty is used for declaring the water and hydrocarbon flows is typically less than 300 kPa (gage)
calibration and measurement capability (CMC). On the other hand, the considering the specification of a commercial pump. However, the
uncertainty of relative deviations is not explained by simply combining pressure range is not very critical because water or hydrocarbons as
the system uncertainty and the uncertainty of DUT. Uncertainty factors working fluid are incompressible fluids. 18
 INTERNATIONAL JOURNAL OF PRECISION ENGINEERING AND MANUFACTURING Vol. 18, No. 8 AUGUST 2017 / 1077

Fig. 1 Schematic diagram of master meter method

2.2 Procedure for comparison calibration Here, E is the relative deviation (%), m· DUT is the mass flow rate of
A procedure for the comparison calibration can be summarized as DUT (kg/s), and m· REF is the mass flow rate of REF (kg/s). Since the
follows: mass flow rate consists of volume flow rate and density, Eq. (1) can be
1) Check the resolution of the output signals of the REF and the written as follows.
DUT.
ρDUT qDUT
2) Install the REF and the DUT along the pipeline which is suitable E = ----------
- ----------- – 1 (2)
ρREF qREF
for flow range of the REF and the DUT.
3) Adjust zero calibrations to the REF and the DUT. Here, ρDUT is the liquid density of DUT (kg/m ), ρREF is the liquid 3

4) Check the range and resolution of output signals from the density of REF (kg/m ), qDUT is the volume flow rate of DUT (m /h),
3 3

associated instruments such as a multi-meter and a pulse counter. and qREF is the volume flow rate of REF (m /h). If liquid temperature
3

5) Turn on the pump and operate the flow meter calibration system is stable enough to guarantee the temperature measurement within the
to pre-heat all the measurement instruments. Operate the flow system accuracy as described in the section 2.1, liquid density variation can be
at the maximum flow rate more than 30 minutes. neglected. This can make Eq. (2) be simplified as follows.
6) Remove air bubbles along the pipeline at the maximum flow rate.
qDUT
7) Check if all the output signals such as pulses, electric currents, E = ----------
-–1 (3)
qREF
temperature and pressure are normal by operating the flow control
valves attached to the pipeline in the flow meter calibration system. By the way, the uncertainty of E, i.e., u(E), can be decomposed into
8) Make sure that the flow rate is in steady state whenever the flow type A evaluation of uncertainty, which is obtained by repeatable
rate is changed. Time delay due to the change of open area ratio in the measurements, and type B evaluation of uncertainty, which is traceable
7,8
flow control valve must be considered. to the SI units.
9) Record the output signals of the REF and the DUT.
2 2
10) Record temperature change of the working fluid at the two u (E ) = u (EA ) + u (EB ) (4)
locations where the REF and the DUT are located. Check the constant
density assumption during the calibration. Here, u(EA) is the type A evaluation of u(E) (%), which can be defined
7,8
11) Repeat the procedure 8)-10). Measurement points should be as follows.
more than 5 locations by changing flow rates. At each measurement
1 n 2
point, measurement should be repeated more than 3 times as the u ( EB ) = ------------------ Σi =1( Ei – E ) (5)
n(n – 1)
following measurement plan.
i) If repeatable measurement is only considered, flow measurement u(EB) is the type B evaluation of u(E) (%) expressed as follows.
is performed from the maximum flow rate to the minimum flow rate,
2 2
or vice versa. At each flow rate, consecutive measurements are u ( EB ) = ( cqDUT u (δ qDUT ) ) + ( cqREF u (δ qREF ) ) (6)
performed.
ii) If the effect of hysteresis is considered, flow measurement is Here, δqDUT is the correction value of qDUT, u(δqDUT) is the type B
performed as follows: From the maximum to the minimum flow rate, evaluation of the uncertainty of qDUT, δqREF is the correction value of
and from the minimum to the maximum flow rate, and from the qREF, and u(δqREF) is the type B evaluation of the uncertainty of qREF.
maximum to the minimum flow rate, etc.. cqDUT is the sensitivity coefficient of E to qDUT, cqREF is the sensitivity
12) Make a calibration report by calculating relative deviations. coefficient of E to qREF, u(qDUT) is the uncertainty of DUT, and u(qREF)
is the uncertainty of REF. cqDUT and cqREF can be calculated by first-
7,8
2.3 Mathematical model order partial differentiation as follows.
The mathematical model for comparison calibration between the
∂u ( E ) 1 E+1
REF and the DUT is as follows. cqDUT = -------------- = ---------- = -----------
∂qDUT qREF qDUT
(7)
m· DUT – m· REF ∂u ( E ) qDUT E+1
E = ----------------------------
- (1) cqREF = -------------- = –-----------
- = –-----------
m· REF ∂qREF qREF
2 qREF
 1078 / AUGUST 2017 INTERNATIONAL JOURNAL OF PRECISION ENGINEERING AND MANUFACTURING Vol. 18, No. 8

Relative uncertainty can be defined as ur as follows. relative uncertainty due to the long-term stability (%). All the terms can
be explained as follows.
u (δ qDUT ) u (δ qREF )
ur (δ qDUT ) = ---------------------
-, ur (δ qREF ) = --------------------
- (8)
qDUT qREF
2.5.1 Associated instrument
Relative sensitivity can be also defined as cr as follows. ur(δqREF ) is the relative uncertainty of resolution of an associated
1

instrument to the REF. It consists of 1) a digital indicator, 2) an analog

cr qDUT = E + 1, cr qREF = –( E + 1 ) (9) indicator, 3) a multi-meter for measuring volume flow rate, or 4) a
pulse counter for measuring fluid volume. The purpose of the items 3)
If Eqs. (7)-(9) are substituted into Eq. (6), u(EB) becomes as follows. and 4) is to relate the REF to the associated instruments which are
traceable to the SI units.
2 2
u ( EB ) = ( cr qDUT u r (δ q DUT ) ) + ( cr qREF u r (δ qREF ) ) (10)
2.5.1.1 Digital indicator
If Eq. (10) is substituted into Eq. (4), u(E) becomes as follows. If a digital indicator is used, ur(δqREF ) becomes as follows.
1

2 2 2 1 δ qmin
u ( E ) = u (E A ) + ( cr qDUT u r (δ q DUT ) ) + ( cr qREF u r (δ q REF ) ) (11) ur (δ qREF 1 ) = ---------- -----------
- (14)
2 3 qREF

This is the equation for evaluating the combined standard uncertainty Here, δqmin is the resolution of indicated output of REF (m /h; m ). The 3 3

of E according to the GUM and the EA-4/02. It is notable that the type 7,8
half of δqmin follows the rectangular probability distribution. 7,8

A evaluation of uncertainties of u(qREF) and u(qDUT) are incorporated

into u(EA). The purpose of this treatment is to avoid complicated 2.5.1.2 Analog indicator
calculations due to the covariance between qREF and qDUT. If an analog indicator is used, ur(δqREF ) becomes as follows.
1

1 δ qmin
2.4 Type A evaluation of uncertainty of the relative deviations ur (δ qREF 1 ) = ---------- -----------
- (15)
2 6 qREF
All the type A evaluations of uncertainties by qREF and qDUT can be
summarized into the type A evaluation of uncertainty by E. It is because The half of δqmin follows the triangular probability distribution. 7,8

qREF and qDUT are incorporated into E. This can reduce cumbersome
calculations of covariance between qREF and qDUT. The covariance must 2.5.1.3 Multi-meter
be included if the type A evaluation of uncertainty of E is to be If a multi-meter is used, ur(δqREF ) becomes as follows.
1

explained by the type A evaluations of u(qREF) and u(qDUT). u(EA) can 7,8

2 2
be expressed as follows. ur (δ qREF 1 ) = cr iREF ur (δ iREF 1) + ur (δ iREF 2 ) (16)
2 2
( cr qDUT u r ( q DUT A ) ) + ( cr qREF u r ( q REF A ) )
u ( EA ) = Here, cr i is the relative sensitivity coefficient of qREF to iREF, iREF is
REF
+ 2r ( qDUT, qREF )cr qDUT c r qREF u r ( q DUT A )u r ( q REF A ) the indicated output of the multi-meter (mA), ur(δ iREF ) is the relative1
(12)
2 2 uncertainty due to the calibration certificate (%), and ur(δ iREF ) is the
( ( E + 1 )ur ( qDUT A ) ) + ( –( E + 1 )ur ( qREF A ) )
2

= relative uncertainty due to the long-term stability (%). The conversion

2
–2 ( E + 1 ) r ( qDUT, qREF )ur ( qDUT A )ur ( qREF A ) formula between iREF and qREF is useful as follows.
Here, ur(qDUT A) is the relative type A evaluation of u(qDUT) (%),
qmax – qmin
ur(qREF A) is the relative type A evaluation of u(qREF) (%), and r(qDUT, qREF = ----------------------
- ( i – i ) + qmin (17)
imax – imin REF min
qREF) is the cross-correlation between qDUT and qREF. It is notable that
r(qDUT, qREF) is a necessary part to explain the covariance between qREF Here, qmax is the maximum volume flow rate (m /h; m ), qmin is the 3 3

and qDUT according to the GUM and the EA-4/02. If the number of 7,8
minimum volume flow rate (m /h; m ), imax is the maximum electric
3 3

n input variables to the measurands are increased, the number of cross- current (mA), and imin is the minimum electric current (mA). cr iREF
correlations are increased by nC . Thus, it is practical to estimate the
2 can be complicated if the first-order partial differentiation of qREF with
type A evaluation of uncertainty of the measurand instead of considering respect to iREF is applied. However, it is more practical to set cr iREF as
all the type A evaluations of uncertainties for the input variables. 1 because cr iREF is the relative sensitivity coefficient.

2.5 Type B evaluation of uncertainty of the REF 2.5.1.4 Pulse counter

The relative combined uncertainty of corrections to the output of If a pulse counter is used, ur(δ qREF ) becomes as follows.
1

REF, δqREF, is expressed as follows.

2 2
ur (δ qREF 1 ) = cr fREF ur (δ fREF 1) + ur (δ fREF 2 ) (18)
2 2 2
ur (δ qREF ) = ur (δ qREF 1 ) + ur (δ qREF 2 ) + ur (δ qREF 3 ) (13)
Here, cr f is the relative sensitivity coefficient of qREF to fREF, fREF is
REF
Here, ur(δqREF ) is the relative uncertainty due to associated instrument
1 the indicated output of the pulse counter (pulse), ur(δ fREF ) is the 1

(%), ur(δqREF ) is the relative uncertainty due to the calibration

17
2 relative uncertainty due to the calibration certificate (%), and ur(δ fREF ) 2

certificate which is traceable to the SI units (%), and ur(δqREF ) is the 3 is the relative uncertainty due to the long-term stability (%). The
 INTERNATIONAL JOURNAL OF PRECISION ENGINEERING AND MANUFACTURING Vol. 18, No. 8 AUGUST 2017 / 1079

Fig. 2 Schematic diagram for a water flow standard system

conversion formula between fREF and qREF is as follows. not needed. It is because the comparison calibration should determine
the uncertainties due to calibration certificate and the long-term stability.
fREF
qREF = ------------
- (19) Therefore, the type B uncertainty of the DUT can be expressed as
KtREF
follows.
Here, K is the K-factor (pulse/m ) of REF and tREF is the elapsed time
3

(s) to convert fluid volume into volume flow rate. cr fREF becomes 1 ur (δ qDUT ) = ur (δ qDUT 1 ) (22)
since it is the relative sensitivity coefficient.
2.7 Measurement uncertainty
2.5.2 Calibration certificate The measurement uncertainty of E becomes as follows.
ur(δ qREF ) is the relative uncertainty which can be calculated by the
2

measurement uncertainty of REF according to a calibration certificate. U ( E ) = ku ( E ) (23)

U
ur (δ qREF 2 ) = ------s (20) Here, k can be determined by the confidence interval of 95%. Toward
k
this end, the effective degrees of freedom, vE, should be calculated by
Here, Us is the measurement uncertainty of REF denoted in the the Welch-Satterthwaite formula. 7,8

calibration certificate (%), and k is the coverage factor. 4

u(E)
vE = -------------------------------------------------------------------------------------------------------------
4 4 4 4
-
4
(24)
u ( EA) cr qDUTur (δ qDUT) cr qREF ur (δ qREF )
2.5.3 Long-term stability ---------------- + ------------------------------------------ + -----------------------------------------
v EA vqDUT vqREF
ur(δ qREF ) indicates the long-term stability by accumulating the
2

calibration certificates of REF for a certain period of time. The long- vE is the effective degrees of freedom of E, vqDUT is the degrees of
term stability can be found by tracing the variations of E as the freedom of qDUT, and vqREF is the degrees of freedom of qREF.
following equation.

δE
ur (δ qREF 3 ) = ------ (21) 3. Examples of Comparison Calibration
3

Here, δE is the long-term deviations of E (%). The long-term Measurement uncertainties of a Coriolis mass flow meter (E+H
deviations indicate the change of measurement uncertainty in case of 83F1H, S/N 5604DD02000) and an electro-magnetic flow meter (E+H
the gravimetric flow standards or the maximum difference of E divided 53W1H, S/N A8021B18000) were evaluated. In the comparison
by 3 in case of reference flow meters. The maximum difference of calibration, the duration time was 60 s. The number of data was 60 and
E can be found by accumulating calibration certificates. ur(δ qREF ) is3 the sampling interval was 1 s. The average value of the 60 data was
4,5
assumed to have the rectangular probability distribution. represented as one measurement point. Therefore, each measurement
point followed the Gaussian probability distribution. It was because the
2.6 Type B evaluation of uncertainty of the DUT central limit theorem was applicable to the averaged value of a sample
7,8
The relative uncertainty of the corrections to the indicated output of space and the number of samples should be greater than 30. The
DUT, δ qDUT, can be related to the uncertainty due to associated number of samples was more important than the duration time in these
instruments as explained in the Section 2.5.1. The other uncertainty examples. It was because the turbulent time scales should not affect on
factors due to the calibration certificate and the long-term stability are measuring the averaged flow rate in the steady pipe flow.
 1080 / AUGUST 2017 INTERNATIONAL JOURNAL OF PRECISION ENGINEERING AND MANUFACTURING Vol. 18, No. 8

3 3
Table 1 Calibration data of a Coriolis mass flow meter by gravimetric flow standard (40.4 m /h and 200.0 m /h)
iDUT qDUT qREF E E U(E) iDUT qDUT qREF E E U(E)
3 3 3 3
(mA) (m /h) (m /h) (%) (%) (%) (mA) (m /h) (m /h) (%) (%) (%)
6.59357 40.525 40.431 0.23 16.81765 200.276 199.861 0.21
6.59637 40.568 40.485 0.21 16.82915 200.455 200.016 0.22
6.59225 40.504 40.427 0.19 0.21 0.093 16.83655 200.571 200.121 0.22 0.20 0.076
6.59286 40.513 40.436 0.19 16.82918 200.456 200.114 0.17
6.59124 40.488 40.403 0.21 16.79615 199.940 199.614 0.16

3
Table 2 Uncertainty budget of the Coriolis mass flow meter at 40.4 m /h
Xi xi u(xi) p(xi) ci ciu(xi) vi
δqDUT 0.0 m /h3
0.031 % normal 1.0021 0.031 % ∝
δqDUT 1 0.0 m /h3
0.031 % normal 1 0.031 % ∝
δiDUT 1 0.0 mA 0.030 % normal 1 0.030 % ∝
δiDUT 2 0.0 mA 0.008 % rectangular 1 0.008 % ∝
δqREF 0.0 m /h3
0.034 % normal -1.0021 0.034 % ∝
δqREF 2 0.0 m /h3
0.030 % normal 1 0.030 % ∝
δqREF 3 0.0 m /h3
0.015 % rectangular 1 0.015 % ∝
EA 0.21 % 0.007 % t 1 0.007 % 4
E 0.21 % 0.046 % 7293
U(E) 0.093 %

3
Table 3 Uncertainty budget of the Coriolis mass flow meter at 200.0 m /h
Xi xi u(xi) p(xi) ci ciu(xi) vi
δqDUT 0.0 m /h3
0.012 % normal 1.0020 0.012 % ∝
δqDUT 1 0.0 m /h3
0.012 % normal 1 0.012 % ∝
δiDUT 1 0.0 mA 0.012 % normal 1 0.012 % ∝
δiDUT 2 0.0 mA 0.003 % rectangular 1 0.003 % ∝
δqREF 0.0 m /h3
0.034 % normal -1.0020 0.034 % ∝
δqREF 2 0.0 m /h3
0.030 % normal 1 0.030 % ∝
δqREF 3 0.0 m /h3
0.015 % rectangular 1 0.015 % ∝
EA 0.20 % 0.013 % t 1 0.013 % 4
E 0.20 % 0.038 % 316
U(E) 0.076 %

3.1 Calibration of Coriolis mass flow meter by WFSS Table 2 shows the uncertainty budget of the DUT at qmin. The
The Coriolis mass flow meter (the DUT in this example) was measurement uncertainty was evaluated to be 0.093% (k = 2) at qmin.
calibrated by the water flow standard system (WFSS) at KRISS, the The uncertainty contributions by the traceability to the SI units (δqREF)
REF in this example, as shown in Fig. 2. The WFSS consisted of a and the relative uncertainty of associated instruments (δqDUT) were
reservoir, a pump, a constant level head tank, a set of flow control valves, 0.034% and 0.031%, respectively. The relative uncertainty due to
6
a flow diverter, and a weighing tank. The flow range of the WFSS was repeatable measurements (EA) was 0.007%. This value was small
3 3 3
400-2000 m /h (WFSS1), 40-400 m /h (WFSS2), 4-60 m /h (WFSS3), compared with the other uncertainty factors. The reason why the relative
3
and 0.6-6 m /h (WFSS4). The CMC of WFSS has been evaluated as uncertainty of δqDUT was 0.031% was because the averaged electric
6
0.06% (k = 2) in all the flow ranges. Previously, the CMC of WFSS current of the DUT was 6.593 mA. This means that the relative
3
was evaluated as 0.09% (k = 2). Among the four gravimetric flow uncertainty of the correction to the electric current (δiDUT ) was 0.002
1

standards, the WFSS2 was selected to calibrate the DUT which was mA / 6.593 mA = 0.030%. Here, the standard uncertainty of the
installed at a pipeline with diameter of 100 mm. associated instrument (Agilent 34401A, S/N MY47048540) was 0.002
The DUT had a flow computer which could measure volume flow mA (k = 1) according to its calibration certificate. In addition, the long-
rate as well as mass flow rate. It was because the DUT could measure term stability of the associated instrument was 0.0005 mA. Thus, the
11,21
density according to the international standards. As for the relative uncertainty of the corrections by the long-term stability (δiDUT )2

calibration conditions, water temperature was changed between 14.8°C was 0.0005 mA / 6.593 mA = 0.008%.
3
and 15.1°C. Therefore, water density was changed from 999.20 kg/m In a similar manner, the measurement uncertainty was evaluated to
3
to 999.24 kg/m . This satisfied the calibration condition as stated in the be 0.076% (k = 2) at qmax as shown in Table 3. The uncertainty
19
section 2.1. Table 1 shows some of calibration data at two measurement contribution due to the traceability to the SI units (δqREF) was biggest
points (qmin, qmax) among five measurement points as suggested in the among all the uncertainty factors. The uncertainty contribution by the
calibration procedure 11.i of the section 2.2. E at qmin and qmax was relative uncertainty of δiDUT was reduced to 0.002 mA / 16.822 mA =
1

0.21% and 0.20%. 0.012% at qmax. The relative uncertainty of the corrections by the long-
 INTERNATIONAL JOURNAL OF PRECISION ENGINEERING AND MANUFACTURING Vol. 18, No. 8 AUGUST 2017 / 1081

Fig. 3 Temperature and water density variations during calibrating an electro-magnetic flow meter at 204.3 m /h (T and ρ are for the location of
3
1 1

REF, T and ρ are for the location of DUT)

2 2

3 3
Table 4 Calibration data of an electro-magnetic flow meter by a Coriolis mass flow meter (20.7 m /h and 204.3 m /h)
iDUT qDUT qREF E E U(E) iDUT qDUT qREF E E U(E)
3 3 3 3
(mA) (m /h) (m /h) (%) (%) (%) (mA) (m /h) (m /h) (%) (%) (%)
5.31967 20.62 20.65 -0.15 17.06076 204.07 204.38 -0.15
5.32132 20.65 20.67 -0.11 17.04452 203.82 204.08 -0.13
5.32486 20.70 20.72 -0.12 17.07190 204.25 204.38 -0.07
5.32468 20.70 20.72 -0.08 17.05350 203.96 204.18 -0.11
5.32254 20.66 20.68 -0.05 17.06699 204.17 204.41 -0.12
-0.11 0.199 -0.11 0.169
5.32468 20.70 20.73 -0.14 17.06503 204.14 204.23 -0.04
5.32169 20.65 20.67 -0.09 17.05707 204.02 204.28 -0.13
5.32503 20.70 20.73 -0.15 17.06206 204.09 204.27 -0.09
5.32594 20.72 20.74 -0.09 17.05547 203.99 204.24 -0.12
5.32532 20.71 20.74 -0.13 17.05195 203.94 204.15 -0.11

3
Table 5 Uncertainty budget of the electro-magnetic flow meter at 20.7 m /h
Xi xi u(xi) p(xi) ci ciu(xi) vi
δqDUT 0.0 3
m /h 0.039 % normal 0.9989 0.039 % ∝
δqDUT 1 0.0 3
m /h 0.039 % normal 1 0.039 % ∝
δiDUT 1 0.0 mA 0.038 % normal 1 0.038 % ∝
δiDUT 2 0.0 mA 0.009 % rectangular 1 0.009 % ∝
δqREF 0.0 3
m /h 0.091 % normal -0.9989 0.091 % ∝
δqREF 1 0.0 3
m /h 0.039 % normal 1 0.039 % ∝
δiREF 1 0.0 mA 0.038 % normal 1 0.038 % ∝
δiREF 2 0.0 mA 0.009 % rectangular 1 0.009 % ∝
δqREF 2 0.0 3
m /h 0.034 % normal 1 0.034 % ∝
δqREF 3 0.0 3
m /h 0.075 % rectangular 1 0.075 % ∝
EA -0.11 % 0.010 % t 1 0.010 % 9
E -0.11 % 0.099 % 84794
U(E) 0.199 %

term stability (δiDUT ) was also reduced to 0.0005 mA / 16.822 mA =

2 The DUT was calibrated at two flow rates (qmin and qmax) with ten
0.003%. In addition, the relative uncertainty due to the long-term repeatable measurements. qmin was 20 m /h because the electric current
3

stability of WFSS2 (δqREF ) was estimated as 0.015% because the

3 setting could be adjusted to measure volume flow rate lower than 40
3
relative standard uncertainty of the WFSS2 (δqREF ) was reduced from
2 m /h. Temperature and density variations were monitored during
0.045% (k =1) to 0.030% (k = 1). Therefore, the measurement calibration as shown in Fig. 3. The water temperature was changed
uncertainty of the DUT could be declared as 0.10% (k = 2) in this from 19.15°C to 19.25°C. The water density was changed between
3 3
example. 998.48 kg/m and 998.50 kg/m . This satisfied the calibration condition
within 0.2°C and ± 0.04 kg/m as indicated in the section 2.1. E at qmin
3

3.2 Calibration of electro-magnetic flow meter by Coriolis mass and qmax was -0.11% in Table 4.
flow meter Table 5 shows the uncertainty budget of the DUT at qmin. The
The electro-magnetic flow meter (the DUT in this example) was measurement uncertainty was evaluated to be 0.199% (k = 2) at qmin.
calibrated by the Coriolis mass flow meter (the REF in this example). The uncertainty contribution of δqREF was 0.091%. This relative
 1082 / AUGUST 2017 INTERNATIONAL JOURNAL OF PRECISION ENGINEERING AND MANUFACTURING Vol. 18, No. 8

3
Table 6 Uncertainty budget of the electro-magnetic flow meter at 204.3 m /h
Xi xi u(xi) p(xi) ci ciu(xi) vi
δqDUT 0.0 3
m /h 0.012 % normal 0.9989 0.012 % ∝
δqDUT 1 0.0 3
m /h 0.012 % normal 1 0.012 % ∝
δiDUT 1 0.0 mA 0.012 % normal 1 0.012 % ∝
δiDUT 2 0.0 mA 0.003 % rectangular 1 0.003 % ∝
δqREF 0.0 3
m /h 0.083 % normal -0.9989 0.083 % ∝
δqREF 1 0.0 3
m /h 0.012 % normal 1 0.012 % ∝
δiREF 1 0.0 mA 0.012 % normal 1 0.012 % ∝
δiREF 2 0.0 mA 0.003 % rectangular 1 0.003 % ∝
δqREF 2 0.0 3
m /h 0.034 % normal 1 0.034 % ∝
δqREF 3 0.0 3
m /h 0.075 % rectangular 1 0.075 % ∝
EA -0.11 % 0.010 % t 1 0.010 % 9
E -0.11 % 0.085 % 41102
U(E) 0.169 %

uncertainty consisted of the relative uncertainty of the associated it consisted of several steps such as 1) an appropriate assumption of
instrument (δqREF ) of 0.039%, the relative uncertainty of the REF
1 calibration condition, 2) a calibration procedure, 3) a suitable
(δqREF ) of 0.034% and the relative uncertainty by the long-term
2 mathematical model, and 4) uncertainty analysis of the mathematical
stability (δqREF ) of 0.075%. It is notable that the relative standard
3 model. Two examples of the comparison calibration indicated that the
uncertainty of δqREF was evaluated as (0.046 − 0.031 ) % = 0.034%.
1
2 2 0.5
measurement uncertainty of the flow meters could be summarized into
Here, the relative standard uncertainty of the REF was 0.046% (k = 1) uncertainty budgets. When a Coriolis mass flow meter was calibrated
in the section 3.1. It was because the relative uncertainty of the associated with the water flow standard system at KRISS by the gravimetric
instrument of 0.031% in the section 3.1 should be replaced with that of method, the measurement uncertainty of the Coriolis mass flow meter
δqREF , resulting in the relative uncertainty of (0.034 + 0.039 ) % =
1
2 2 0.5
was evaluated to be 0.10% (k = 2). When an electro-magnetic flow
0.052%. This is plausible because the relative uncertainty is bigger than meter was calibrated with the Coriolis mass flow meter by the master
the relative standard uncertainty of the REF. The long-term stability of meter method, the measurement uncertainty was evaluated to be 0.20%
δqREF was changed from 0.08% to 0.21% in the successive calibrations
2 (k = 2). This means that the master meter method can be used to
(09/2015-07/2016). Thus, the relative uncertainty due to δqREF was 3 calibrate flow meters for custody transfer and for regulations of potable
estimated as (0.21 − 0.08)/3 % = 0.075%.
0.5
water with good accuracy as long as the traceable chain between the
In addition, the uncertainty contribution of δqDUT was 0.039%. The gravimetric method and the master meter method is carefully managed.
relative uncertainty of the other associated instrument (Agilent 34401A,
S/N MY47047284) (δiDUT ) was 0.002 mA / 5.325 mA = 0.038%. The
1

relative uncertainty due to the long-term stability of the associated ACKNOWLEDGEMENT

instrument (δiDUT ) was 0.0005 mA / 5.325 mA = 0.009%. The relative
2

sensitivity coefficient of E was ± 0.9989, slightly smaller than 1. This paper was supported by a research project at KRISS in the
However, the relative sensitivity coefficient did not affect the uncertainty Republic of Korea. The title of project is “Measurement and Control
contributions of δqREF and δqDUT. Technology for Continuum Dynamics” (grant No. 17011008).
Similarly, the measurement uncertainty was evaluated to be 0.169%
(k = 2) at qmax as shown in Table 6. Again, the uncertainty contribution
of qmax was 0.083% with the relative uncertainty of the associated REFERENCES
instrument, 0.002 mA / 17.073 mA = 0.012%, and its long-term stability,
0.0005 mA / 17.073 mA = 0.003%. Therefore, the measurement 1. KOLAS, “Calibration Laboratories,” http://www.kolas.go.kr/%20usr/
uncertainty of the DUT could be declared as 0.20% (k = 2) in this gud/cri/CrrcInsttRcognSttusList.do (Accessed 7 JUL 2017)
example.
2. ISO 4185, “Measurement of Liquid Flow in Closed Conduits −
Weighing Method,” 1980.
4. Conclusions 3. BIPM, “Calibration and Measurement Capabilities Mass and Related
Quantities,” http://kcdb.bipm.org/appendixC/country_list.asp?Service
A detailed procedure for evaluating the measurement uncertainty of
=M/FF.9 (Accessed 5 JUL 2017)
water flow meters by the master meter method was explained
completely in this study. Relative deviations were used as a measurand 4. Furuichi, N., Sato, H., Terao, Y., and Takamoto, M., “A New
and they were obtained by comparing the output signals between a Calibration Facility for Water Flowrate at High Reynolds Number,”
reference flow meter (REF) and a test flow meter (DUT). The Flow Measurement and Instrumentation, Vol. 20, No. 1, pp. 38-47,
comparison calibration enabled to measure the relative deviations and 2009.
 INTERNATIONAL JOURNAL OF PRECISION ENGINEERING AND MANUFACTURING Vol. 18, No. 8 AUGUST 2017 / 1083

5. Engel, R., Beyer, K., and Baade, H.-J., “Design and Realization of Guidance to the Selection, Installation and Use of Coriolis
the High-Precision Weighing Systems as the Gravimetric References Flowmeters (Mass Flow, Density and Volume Flow Measurements),”
in PTB’s National Water Flow Standard,” Measurement Science and 2015.
Technology, Vol. 23, No. 7, Paper No. 074020, 2012.

6. Chun, S., Yoon, B.-R., Lee, Y.-B., and Choi, H.-M., “Development
of Water Flow Standard System for Calibrating Water Flow Meters
3
Up to 2000 m /H in KRISS,” FLOMEKO, 2016.

7. ISO/IEC Guide 98-3, “Uncertainty of Measurement − Part 3: Guide

to the Expression of Uncertainty in Measurement (GUM:1995),”
2008.

8. Evaluation of the Uncertainty of Measurement in Calibration,

“Evaluation of the Uncertainty of Measurement in Calibration,” EA-
4/02 M, 2013.

9. ANSI/API, “Orifice Metering of Natural Gas and Other Related

Hydrocarbon Fluids − Concentric, Square-Edged Orifice Meters −
Part 1: General Equations and Uncertainty Guidelines,” in: Manual
of Petroleum Measurement Standards, Chap. 14.3.1, American
Petroleum Institute, 2012.

10. AGA Report No. 9, “Measurement of Gas by Multipath Ultrasonic

Meters,” 2007. http://www.browntechnical.org/content/PDF/
previews-AGA_XQ0701_pre.pdf (Accessed 5 JUL 2017)

11. API MPMS 5.6, “Measurement of Liquid Hydrocarbons,” 2013.

12. API MPMS 5.8, “Measurement of Liquid Hydrocarbons by

Ultrasonic Flow Meters Using Transit Time Technology,” 2014.

13. ISO 17089-1, “Measurement of Fluid Flow in Closed Conduits −

Ultrasonic Meters for Gas − Part 1: Meters for Custody Transfer and
Allocation Measurement,” 2010.

14. ISO 12242, “Measurement of Fluid Flow in Closed Conduits −

Ultrasonic Transit-Time Meters for Liquid,” 2012.

15. OIML R 137-1 & 2, “NMI R 137 Gas Meters − Part 1: Metrological
and Technical Requirements, Part 2: Metrological Controls and
Performance Tests,” 2012.

16. ISO4064-1, “Water Meters for Cold Potable Water and Hot Water −
Part 1: Metrological and Technical Requirements,” 2014.

17. Wright, J., “WGFF Guidelines for CMC Uncertainty and Calibration
Report Uncertainty,” http://www.bipm.org/utils/en/pdf/ccm-wgff-
guidelines.pdf (Accessed 10 JUL 2017)

18. Kundu, P. K., Cohen, I. M., and Dowling, D. R., “Fluid Mechanics,”
Academic Press, 6th Ed., 2015.

19. Tanaka, M., Girard, G., Davis, R., Peuto, A., and Bignell, N.,
“Recommended Table for the Density of Water between 0°C and
40°C Based on Recent Experimental Reports,” Metrologia, Vol. 38,
No. 4, pp. 301-309, 2001.

20. Park, J. T. and Chang, K.-H., “Application of Statistics to Oil Density

Measurement,” Measurement, Vol. 30, No. 2, pp. 95-103, 2001.

21. ISO 10790, “Measurement of Fluid Flow in Closed Conduits −