Sunteți pe pagina 1din 12

INTRODUCTION

This paper provides a guide to estimating the un-


certainty of a flow measurement made with field
grade instruments. It is applicable to calibrated or
uncalibrated turbine or ultrasonic based measure-
ments. Pressure and temperature measurements con-
tribute uncertainty, the uncertainty components of
each instrument are described. A set of typical
manufacturers specifications are used as an example.
The effects of ambient temperature on both mea-
surements are described.
A flow computer acquires signals from the pres-
sure and temperature transducers and converts the
readings into engineering units. A gas chromatograph
and the equation of state contribute uncertainty to
the determination of compressibility under flowing
base conditions. The components that contribute un-
certainty are described and typical values are pro-
posed.
The process of combining uncertainty components
is illustrated based on hypothetical values. The con-
cepts of sensitivity coefficient, expanded uncertainty
and coverage factor are described. The data are or-
ganized such that the relative contributions of the
components can be readily compared. The value of
this organizational structure is illustrated by compar-
ing calibrated and uncalibrated turbine meters.
The paper concludes with a brief discussion of
advanced topics. These include topics from flow
measurement, instrument calibration and measure-
ment uncertainty.
Flow Measurement Uncertainty - Analysis Based on Field Grade Components
Thomas Kegel
Colorado Engineering Experiment Station, Inc., (CEESI)
Nunn, Colorado USA
MEASUREMENT UNCERTAINTY
When a measurement is made there are two im-
portant values associated with the result of the mea-
surement process. The first is the numerical value of
the variable being measured, the second is the un-
certainty associated with that numerical value. This
section provides an introduction to measurement un-
certainty by briefly describing the paper standards
and terminology.
Paper Standards
The uncertainty analysis procedure commonly used
with flow measurement processes is described in the
standards ANSI/ASME MFC-2M
1
, ANSI/ASME
PTC 19.1
2
, and ISO 5168
3
. All three of these stan-
dards are based on an uncertainty analysis method
developed for comparing results of rocket engine
tests
4
.
A document more recently published by ISO de-
scribes an uncertainty analysis procedure originally
developed to compare the test data of different cali-
bration laboratories
5
. From a practical point of view,
the newer ISO method does not differ significantly
from the older method of Ref. 1-3. It has been
adopted by the National Institute of Standards and
Technology (NIST) as well as a number of large cor-
porations as a formal policy
6
. In addition, the latest
revision of the standard ANSI/PTC 19.1
7
includes
the ISO method. It is likely that future revisions of
flow measurement uncertainty standards will adopt
the ISO method. This paper therefore teaches the
application of that method.
Terms and Definitions
In this section are listed some concepts that need
to be defined prior to presentation of the uncertainty
analysis method.
A measurement process consists of the instru-
ments, operators and procedures that result in
the determination of a numerical value for a vari-
able.
The true value is the value determined with a
perfect measurement process. The true value is
always unknown because all measurement pro-
cesses are imperfect to some degree.
The error is the difference between the mea-
sured value and the true value. The error is al-
ways unknown because the true value is always
unknown.
Uncertainty is an estimate of the interval bound-
ing the measured value within which the true
value lies.
An accuracy specification usually means inac-
curacy but can be interpreted as uncertainty. A
manufacturer states that the accuracy is 1.0%
really means that the accuracy is 99%.
Confidence level (or confidence interval) is the
degree of confidence, expressed as a percent,
that the true value lies within the stated uncer-
tainty. A proper uncertainty statement would read:
"q
m
=500 cfh 1.0% at a 95% level of confidence".
This means that 95 out of every 100 observa-
tions are between 495 and 505 cfh.
Within this paper the term percent (%) means
percent of reading unless it is specifically stated
to be percent of full scale.
ANALYSIS PROCEDURE
The simplified analysis procedure consists of the
five basic steps listed below. Each step is described
in detail within the body of the paper.
1. Write an equation, called the data reduction equa-
tion, that gives the desired output as a function of
one or more variables (components).
2. Identify those components of the data reduction
equation that potentially contribute uncertainty.
3. Determine the sensitivity coefficients for each
component in Item 2.
4. Obtain numerical values for the uncertainty of
each component in Item 2.
5. Combine the numerical values obtained in Item 4
to give a numerical value for the uncertainty.
Data Reduction Equation
At a basic level the data reduction equations for
ultrasonic and turbine meters are the same. That equa-
tion is:
!
!
"
#
$
$
%
&
!
!
"
#
$
$
%
&
!
!
"
#
$
$
%
&
'
f
b
f
b
b
f
f b
Z
Z
T
T
P
P
V V
[1]
where:
V = volume
P = pressure
T = temperature
Z = compressibility
and the subscript b and f refer to base and flowing
conditions.
Uncertainty Components
The following variables are present in Eq. 1:
P
b
, P
f
, T
b
, T
f
, Z
b
, Z
f
, and V
f
Each variable has the potential to contribute un-
certainty.
Sensitivity Coefficients
A sensitivity coefficient must be determined for
each of the variables that contribute uncertainty to
Eq. 1. The sensitivity coefficients are required when
the components of uncertainty are combined at the
end of the analysis procedure. Given an output y that
is a function of n variables:
) ,... , , (
3 2 1 n
x x x x f y ' [2]
The sensitivity of y to x
i
is given by:
!
!
"
#
$
$
%
&
!
!
"
#
$
$
%
&
(
(
'
y
x
x
y
S
i
i
xi [3]
From a practical standpoint, the sensitivity coeffi-
cient can be interpreted as the percent change in y
that results from a one percent shift in x
i
. In the spe-
cific case of Eq. 1 the sensitivity coefficient absolute
values are all unity:
S
Vf
= S
Pf
= S
Tb
= S
Zb
= S
Pb
= S
Tf
= S
Zf
meaning a 1% change in any variable will result in a
1% change in V
b
. The orifice meter equation is more
complex than Eq. 1, the sensitivity coefficients are
likewise more complex. A detailed description is con-
tained in Reference 8.
Numerical Uncertainty Values
The objective in this section is to first identify and
then classify the numerical values of uncertainty. In
order to combine the uncertainties in the individual
components they have to be defined in a uniform
manner. The standard uncertainty
5
is the term given
to the uniform method of expressing numerical val-
ues of uncertainty.
In order to estimate a standard uncertainty value,
a component uncertainty is classified based on how
the numerical value is determined. An estimate is clas-
sified as Type A when statistical data are available.
The standard uncertainty is defined as the statisti-
cally determined standard deviation
9
. An estimate is
classified as Type B when statistical data are not
available. The standard uncertainty is defined as u =
0.58U where U defines the limits within which the
true value is expected to lie. In some cases the stan-
dard uncertainty of a Type B estimate is u = 0.50U
if it is known to be based on a statistical analysis.
Most of the standard uncertainties in the present ex-
ample are classified as Type A. The use of flowme-
ter calibration data is discussed at the end of the pa-
per, this process illustrates the use of Type B esti-
mates.
The pressure and temperature values that define
base conditions (P
b
and T
b
) are fixed and contribute
no uncertainty. The important consideration is that
consistent values of P
b
and T
b
are maintained. The
values of flowing conditions (P
f
, T
f
, Z
f
, and V
f
) are
determined from measurements and therefore con-
tribute uncertainty to the value of V
b
.
The uncertainty in base compressibility is a bit more
complex. On one hand it is based on fixed values of
P
b
and T
b
and would contribute no uncertainty. On
the other hand it is based on gas composition which
requires a measurement and therefore contributes
uncertainty.
The uncertainty analysis process is best described
based on numerical values. A hypothetical set of mea-
surements has been developed to provide these val-
ues. They are not intended to represent an actual
measurement. The details of determining the hypo-
thetical values are described in the next four sections,
the resulting values are summarized below:
u
Pb
=u
Tb
=0
u
Pf
= 2.73 psi = 0.26 %
u
Tf
= 0.73 F = 0.14 %
u
Zf
= 0.20 %
u
Zb
= 0.06 %
u
Vf
= 0.58% (turbine meter)
= 0.41% (ultrasonic meter)
Uncertainty components can be expressed in en-
gineering units (ft
3
, psia, F) or as percent of mean
value. Using percentages allow for comparison be-
tween components and simplifies the math. It is em-
phasized that components must be expressed as per-
cent of reading, not percent of full scale.
Combining Uncertainty Components
The individual component standard uncertainty
values are combine to result in the combined stan-
dard uncertainty
5
. The combining equation is:
) *
+
'
'
n
i
xi xi y
u S u
1
2 2
[4]
In the present example Eq. 4 is written as:
2 2 2 2 2 2
Vf Zb Zf Tf Pf Vb
u u u u u u , , , , ' [4a]
Combining the numerical values:
2 2 2 2 2 2
58 . 0 06 . 0 20 . 0 14 . 0 26 . 0 , , , , '
Vb
u
results in u
Vb
= 0.68 % for the turbine meter, and
2 2 2 2 2 2
41 . 0 06 . 0 20 . 0 14 . 0 26 . 0 , , , , '
Vb
u
results in u
Vb
= 0.54 % for the ultrasonic meter.
The expanded uncertainty
5
is the product of stan-
dard uncertainty and a coverage factor. That prod-
uct defines the confidence level. Most measurements
are stated at 95% confidence with a coverage factor
of 2.0. In the present example the expanded uncer-
tainties are 2.00.68 % = 1.36 % for the turbine meter
and 2.00.54 % = 1.08 % for the ultrasonic meter.
Using a coverage factor of 2.0 in the present ex-
ample is a simplifying assumption. A different value
may be required in a real uncertainty analysis. The
details represent an advanced topic beyond the scope
of this paper.
PRESSURE MEASUREMENT
The hypothetical pressure transmitter is config-
ured as follows:
Upper Range Limit (URL) = 2500 psig
Span = 1400 psig (20 mA)
Zero = 500 psig (4 mA)
Measured Pressure = 1054 psig (13.85 mA)
The uncertainty in pressure measurement is made
up of six components:
1. Combined Performance Specification
2. Stability
3. Ambient Temperature Effect
4. Calibration Process
5. Barometric Pressure
6. Data Acquisition
Determining numerical values for the six compo-
nents is discussed below. The behavior of four of the
components is shown in Figure 1. The x-axis is pres-
sure, the y-axis is the uncertainty expressed as a per-
cent of reading. The uncertainty decreases as the
pressure increases, this is due to full scale effects.
Such effects are present in all measurement pro-
cesses. In some processes the relative magnitude may
be very small and the effect may not be apparent.
The components of uncertainty for pressure mea-
surement are combined using Eq. 4, the units of u
are psi. Combining the numerical values:
2 2 2 2 2 2
99 . 0 12 . 0 47 . 0 18 . 2 22 . 1 , , , , '
Pf
u
results in u
Pf
= 2.73 psig which is 0.26% of 1068 psia.
The uncertainty must be expressed as a percent of
absolute (psia) not gage (psig) units because the pres-
sure value is used to calculate compressibility.
Combined Performance Specification
In the present example the vendor claims the ac-
curacy is 0.15% of span, the term accuracy is
said to include the four components listed below.
Regardless of vendor terminology, the user should
make sure that all four components are included.
1. Repeatability
2. Reproducibility
3. Linearity
4. Hysteresis
In the present example the span is fixed and so
this uncertainty component can be calculated in units
of psig. Doing so results in:
psig 1400
% 100
% 15 . 0
span % 15 . 0 - !
"
#
$
%
&
. ' . ' U
= 2.10 psig [5a]
u = 0.58(2.10)=1.22 psig [5b]
0.0
0.2
0.4
0.6
0.8
500 700 900 1100 1300 1500
Pressure [psig]
S
t
a
n
d
a
r
d

U
n
c
e
r
t
a
i
n
t
y

[
%
]
Total
Stability
Accuracy
Data Acq
Amb Temp
Figure 1: Pressure Measurement
Uncertainty Components
Repeatability and Reproducibility
Repeatability is the ability of the transmitter to
measure a constant pressure over a short period
of time. Reproducibility is the ability of the trans-
mitter to measure the same pressure subject to varia-
tions that occur over a long period of time. A short
period of time might be measured in minutes while a
long period of time might be months.
Linearity
The sensitivity is the ratio of pressure transmit-
ter input (psi) to output (mA). In the present example
the sensitivity is 56.25 [psi/mA]. The linearity de-
fines the consistency of the sensitivity value over the
pressure range. Stated differently, the linearity mea-
sures the degree to which the input and output are
related by a straight line.
Hysteresis
This effect accounts for a dependence of the out-
put on the direction change of the input. An example
would be the backlash present in a gearing system. A
good machinist will eliminate backlash by first back-
ing out the tool and then moving it into the part. In a
pressure transmitter the hysteresis arises from the
mechanical properties of the sensor. Hysteresis con-
tributes measurement uncertainty because the direc-
tion of change of a measureand prior to a measure-
ment is generally unknown.
Stability
The vendor claims the stability is 0.25% of
URL. Calculating uncertainty in pressure units re-
sults in:
psig 2500
% 100
% 25 . 0
URL % 25 . 0 - !
"
#
$
%
&
. ' . ' U
= 3.75 psig [6a]
u = 0.58(3.75) =2.18 psig [6b]
Ambient Temperature Effect
It has been proposed that all measuring devices
are first thermometers. Every instrument exhibits
sensitivity to changes in ambient temperature. In a
pressure transmitter the ambient temperature effects
the mechanical properties of the elastic element. The
uncertainty component accounts for the magnitude
of the effect. In the present example the vendor states
an uncertainty of:
) * span % 35 . 0 URL % 45 . 0 , . ' U [7]
for every 100F change in the ambient temperature.
In the present example the pressure transducer is
installed in an environment where the temperature
variation is controlled to within 2.5F which repre-
sents a total potential temperature change of 5F. The
uncertainty due to the ambient temperature effect
therefore becomes:
) *
!
!
"
#
$
$
%
&
- , . '
F 100
F 5
span % 35 . 0 URL % 45 . 0
!
!
U
= 0.81 psig [8a]
u = 0.58(0.81) = 0.47 psig [8b]
Calibration Process
In the present example the role of the calibration
process is to maintain the transmitter within the origi-
nal performance specifications. No additional uncer-
tainty is assumed to be contributed by this compo-
nent.
Barometric Pressure
In the present example a fixed value of baromet-
ric pressure is assumed. This value is calculated based
on the elevation of the meter station. It is assumed
that the barometric pressure can vary by 0.2 psia
due to meteorological conditions. This variation rep-
resents an uncertainty in the fixed value
Data Acquisition
The 4-20 mA transmitter output is converted into
engineering units by a flow computer. This process
adds uncertainty in the conversion of the analog sig-
nal to digital. The vendor claims an accuracy of
0.1% of full scale. The term %full scale from
the computer vendor is assumed to mean the same
as %span from the transducer vendor. The con-
version process to pressure units is similar as that
used with Eq. 5:
psig 1400
% 100
% 15 . 0
span % 10 . 0 - !
"
#
$
%
&
. ' . ' U
= 1.71 psig [9a]
u = 0.58(1.71) =0.99 psig [9b]
TEMPERATURE MEASUREMENT
The hypothetical temperature transmitter is con-
figured as follows:
Span = 100 F
Zero = 0 F
Measured Temperature = 50 F (12 mA)
The uncertainty in temperature measurement is
made up of the following components:
1. Combined Performance Specification
2. Stability
3. Ambient Temperature Effect
4. Calibration Process
5. RTD Probe
6. Self Heating
7. Heat Transfer Effects
8. Data Acquisition
The determination of numerical values is described
below. Combining the values:
) *
2 2 2 2 2 2
59 . 0 01 . 0 37 . 0 03 . 0 15 . 0 2 , , , , '
Tf
u
results in u
Tf
= 0.73 F which is 0.14% of 510 R. The
uncertainty must be expressed as a percent of abso-
lute (R) units because the temperature value is used
to calculate compressibility.
Combined Performance Specification
In the present example the vendor claims the ac-
curacy is 0.25% of span:
F 100
% 100
% 25 . 0
span % 25 . 0 - !
"
#
$
%
&
. ' . ' U
= 0.25F [10a]
u = 0.58(0.25)=0.15F [10b]
The term accuracy is assumed to include the
same four components as discussed with the pres-
sure measurement.
Stability
In the present example the vendor claims the sta-
bility is 0.25% of span. The standard uncertainty is u
= 0.15 F.
Ambient Temperature Effect
The vendor states an uncertainty of:
0.25%span) F 4 . 0 ( , . ' U [11]
for every 50F change in the ambient temperature.
In the present example the temperature transducer
is installed in an environment where the temperature
variation is controlled to within 2.5F which repre-
sents a total potential temperature change of 5F. The
uncertainty due to the ambient temperature effect
therefore becomes:
) *
!
!
"
#
$
$
%
&
- , . '
F 50
F 5
span % 25 . 0 F 4 . 0
!
!
U
= 0.05 F [14a]
u = 0.58(0.05) = 0.03 F [14b]
Calibration Process
In the present example the role of the calibration
process is to maintain the transmitter within the origi-
nal performance specifications. No additional uncer-
tainty is assumed to be contributed by this compo-
nent.
RTD Probe
The vendor claims an uncertainty that varies with
temperature. In the current example that value is U
= 0.63 F (u = 0.37 F)
Self Heating
When current flows through a resistor heat is gen-
erated. The heat results in a potential measurement
offset that is estimated as uncertainty. In the present
example this effect is assumed to be included in the
RTD probe uncertainty.
Heat Transfer Effects
The transmitter output indicates the temperature
of the tip of the temperature sensor RTD. The de-
sired value is the temperature of the flowing gas. If
there is a temperature difference between the flow-
ing gas and the pipe wall heat transfer will result.
Heat transfer is present between the RTD and pipe
wall due to conduction through the thermowell as well
as direct radiation. The result is to drive the RTD
temperature closer to the pipe wall temperature. Heat
transfer due to convection is present between the
thermowell and the flowing gas. The convection will
drive the RTD temperature toward that of the flow-
ing gas. The difference between the RTD reading
and the flowing gas temperature (
Tht
) is a measure-
ment error, the magnitude must be estimated as an
uncertainty component.
The numerical value of
Tht
is a function of gas
density, gas velocity, and thermowell geometry as well
as the difference between flowing gas and pipe wall
(
Tgw
)
10
. The relationship between
Tht
and gas ve-
locity is shown in Figure 2 based on the following
conditions:
thermowell: 0.5 OD, 0.25 ID, 3 length within
pipe
pipe wall temperature = 60 F and 80 F
gas temperature = 50 F
gas pressure = 1068 psia
In the present example it is assumed that the gas
velocity is 20 ft/s and the pipe wall temperature is 80
F. Under such conditions
Tht
= 0.02 F (u = 0.01
F).
Data Acquisition
The 4-20 mA transmitter output is converted into
engineering units by the same flow computer as used
for pressure measurement. The standard uncertainty
is u = 0.59 F.
COMPRESSIBILITY
The hypothetical gas composition used in the
present example is contained in column 2 of Table 1.
The process of determining uncertainty in calculated
compressibility is made up of four components:
1. State Equation
2. Chromatograph Repeatability and Reproducibil-
ity
3. Chromatograph Calibration Standard
4. Sensitivity Coefficients
The determination of numerical values is described
below. Combining the values:
2 2 2 2 2 2
002 . 0 06 . 0 , 19 . 0 06 . 0 , ' , '
Zb Zf
u u
result in u
Zf
= 0.20 % and u
Zb
= 0.06 %.
State Equation
The AGA 8 equation of state
11
provides a value
of compressibility given inputs of pressure, tempera-
ture and gas composition. The uncertainty in the state
equation is U = 0.1% (u = 0.06%).
Figure 2: Heat Transfer Effects in
Temperature Measurement
0.1
1
10
1 10
Velocity [ft/s]
T
e
m
p
e
r
a
t
u
r
e

D
i
f
f
e
r
e
n
c
e
,

/
T
h
t

[

F
]
/T
gw
= 30F
/T
gw
= 10F
Chromatograph Repeatability and
Reproducibility
The repeatability and reproducibility of a typical
gas chromatograph is described in the GPA 2261 stan-
dard
12
. The stated values are contained in column 3
of Table 1. The standard does not provide a value for
hydrogen, in the current example the tabulated value
is set equal to the nitrogen value.
Calibration Standard
The chromatograph is calibrated using a gravi-
metric gas standard. The uncertainty (by mass) in a
typical calibration standard is:
U = 1% (u = 0.58%) for components that make
up between 2% and 49% of the mixture
U = 2% (u = 1.16%) for components that make
less than 2% of the mixture
The calibration standard uncertainty values are
contained in column 4 of Table 1. The uncertainty in
the mass of methane (U = 0.25%) is calculated based
on the uncertainties of the other components of the
mixture.
Sensitivity Coefficient
This component accounts for the sensitivity in cal-
culated compressibility to the uncertainty in compo-
sition. In other words, how much uncertainty in cal-
culated compressibility results from the uncertainties
in columns 3 and 4 in Table 1.
The partial derivative in Eq. 4 cannot be deter-
mined by conventional methods because of the com-
plexities in the state equation, a Monte Carlo simula-
tion
13
is used as an alternative. This method is based
on generating multiple unique gas compositions by
randomly varying the mole fraction for each compo-
nent. The standard deviation of a component random
variance is selected to match the combined effect of
the values contained in columns 3 and 4 of Table 1.
The present analysis is based on 2500 simulated
gas compositions. The standard deviation values are
u = 0.19% for Z
f
and u = 0.002% for Z
b
. These val-
ues account for the uncertainties associated with the
chromatograph repeatability and reproducibility and
calibration standard.
t n e n o p m o C n o i t c a r F e l o M [ R d n a R U ] % [ d t S l a C U ] %
e n a h t e M 5 1 0 8 . 0 9 2 . 0 5 2 . 0
e n a h t E 8 8 9 0 . 2 0 . 1 0 . 1
e n a p o r P 0 2 0 0 . 2 0 . 1 0 . 1
e n a t u B - n 3 0 0 9 . 0 0 . 2 0 . 2
e n a t u B - i 1 0 0 9 . 0 0 . 2 0 . 2
e n a t n e P - n 0 0 0 4 . 0 0 . 3 0 . 2
e n a t n e P - i 0 0 0 4 . 0 0 . 3 0 . 2
+ 6 C 0 0 0 2 . 0 0 . 0 1 0 . 2
n e g o r t i N 2 8 9 4 . 1 0 . 2 0 . 2
e d i x o i D n o b r a C 1 9 9 3 . 0 0 . 3 0 . 2
n e g o r d y H 0 0 0 4 . 0 0 . 2 0 . 2
Table 1: Gas Components
VOLUME AT FLOWING CONDITIONS
The volume flow is measured using either a tur-
bine or ultrasonic meter. The uncertainties for using
these meters are specified by the AGA 7
14
and AGA
9
15
standards. For operation at higher flowrates the
uncertainties are 1.0% (u = 0.58%) for the turbine
meter and 0.7% (u = 0.41%) for the ultrasonic
meter. Performance of either meter at lower flowrates
will have higher uncertainties, the low flow condi-
tions are not considered in this example.
REDUCING THE UNCERTAINTY
The uncertainty values for all sixteen components
for turbine meter based measurement are contained
in Table 2. Column 2 shows uncertainty components
expressed as percentages of mean values. Column 3
show the percent contribution to combined uncertainty
made by each component, they are calculated based
on
% 100
2
-
!
!
"
#
$
$
%
&
y
xi
u
u
. The summary in Table 2 is a pow-
erful tool in allocating measurement resources based
on measurement uncertainty. The turbine meter, for
example, is the largest contributor to uncertainty. It
represents the best return on an investment intended
to reduce uncertainty. The ambient temperature ef-
fects, on the other hand, contribute very little uncer-
tainty. The overall uncertainty cannot be reduced by
investing in reduced uncertainty in ambient tempera-
ture effects.
The current example concludes with a hypotheti-
cal calibration of the turbine meter. The calibration
data are contained in Figure 3, the y-axis is percent
shift in K Factor from a nominal values. The data are
well within AGA 7 specifications, the uncertainty can
be reduced as a result. The fitted curve is slightly
nonlinear, the K Factor changes with flowrate. This
will require a more complex flow computer algorithm.
The solid lines in Fig. 3 represent the 95% confidence
interval, 95% of the data are within the interval. The
interval width is 0.2% (u = 0.1%) which accounts
for random effects present during the calibration pro-
cess. The uncertainty of the hypothetical calibration
facility flowrate measurement is 0.25% (u=0.13%)
which accounts for systematic effects associated with
the facility itself. In the current example these two
components make up the uncertainty associated with
the turbine meter measurement. The recalculated un-
certainty is u
Vb
= 0.39%, the details are contained in
Table 3. Having reduced the turbine meter uncer-
tainty, it can be seen that other components now domi-
nate the uncertainty. The gas chromatograph and pres-
sure transducer stability are two examples.
GENERAL DISCUSSION
It is important to emphasize that the analysis pre-
sented above is intended as an example only. Appli-
cable numerical values need to be determined for a
real uncertainty analysis. In addition, several poten-
tial uncertainty components have been neglected in
the interest of simplifying the example. These include,
but are not limited to:
flowmeter installation effects
gas sampling
wet gas
pulsating flow
Considerable analysis was required to estimate
uncertainties of components that ended up contribut-
ing very little uncertainty. Two examples are the heat
transfer and the chromatographic analysis determi-
nation of base compressibility. Unfortunately, the
magnitude of an uncertainty component is unknown
until the analysis is complete.
Figure 3: Turbine Meter Calibration
Data
-0.6
-0.3
0.0
0.3
0.6
20 40 60 80 100
Flowrate [%max]
K

F
a
c
t
o
r

S
h
i
f
t

[
%
]
A number of advanced concepts of uncertainty
analysis are not included in the interest of simplifying
the example. These components can significantly
change the results of an uncertainty analysis, they
cannot be ignored in a real analysis. One concept is
correlated effects
5
where some components of un-
certainty will cancel out if the sensitivity coefficients
are equal but opposite in sign. In some analyses the
presence of correlated components will increase un-
certainty. A valuable tool in advanced uncertainty
analyses is statistical process control
16, 17
. This tool
provides two benefits, it monitors the stability of a
measurement process and identifies sources of ran-
dom effects. Finally, the current example briefly il-
lustrated the turbine meter calibration process. Simi-
lar reductions in uncertainty are possible by imple-
menting calibration processes for the pressure, tem-
perature and gas composition measurements. In gen-
eral calibration replaces a smaller number of Type B
estimates with larger number of Type A estimates.
Calibration generally results in a more complex analy-
sis but lower uncertainty. The complexity arises in
part due to advanced concept of measurement trace-
ability.
REFERENCES
1. ANSI/ASME MFC-2M, Measurement Uncer-
tainty for Fluid Flow in Closed Conduits, Ameri-
can Society of Mechanical Engineers, 1988
2. ANSI/ASME PTC 19.1, Measurement Uncer-
tainty, American Society of Mechanical Engi-
neers, 1990.
3. ISO 5168, Measurement of Fluid Flow - Estima-
tion of Uncertainty of a Flow Rate Measurement,
International Organization for Standardization,
1978
4. Abernethy, R. B. et al, Handbook Uncertainty in
Gas Turbine Measurements, AEDC-TR-73-5,
Arnold Engineering Development Center, 1973
5. ISO Guide to the Expression of Uncertainty in
Measurement, International Organization for
Standardization, 1994
6. Taylor, B. N., and Kuyatt, "Progress Report on
the Implementation of the ISO Guide to the Ex-
pression of Uncertainty in Measurement", Proc.
1994 Meas. Sci. Conf, 1994
7. ANSI/ASME PTC 19.1, Test Uncertainty,
American Society of Mechanical Engineers,
1998.
8. Kegel, Thomas, "Basic Measurement Uncer-
tainty," 74th International School of Hydrocar-
bon Measurement, Tulsa, Oklahoma, May 25-27,
1999.
9. Wadsworth, H. M., Handbook of Statistical Meth-
ods for Engineers and Scientists, McGraw-Hill,
1990.
10. Doebelin, E., O., Measurement Systems: Appli-
cations and Design, McGraw Hill, 1975.
11. Starling, K. E. and Savidge, J. L.,Compressibility
Factors of Natural Gas and Other Related Hy-
drocarbon Gases, Transmission Measurement
Committee Report No. 8, American Gas Asso-
ciation, 1994.
12. GPA 2261 Analysis for Natural Gas and Similar
Gaseous Mixtures by Gas Chromatography, Gas
Processors Association.
13. Basil, M. and Jamieson, A. W., Uncertainty of
Complex Systems by Monte Carlo Simulation,
North Sea Flow Measurement Workshop,
Gleneagles, 26 29, October 1998.
14. ___,Measurement of Fuel Gas by Turbine
Meters, Transmission Measurement Committee
Report No. 7, American Gas Association, 1981.
15. ___,Measurement of Gas by Multipath Ultra-
sonic Meters, Transmission Measurement Com-
mittee Report No. 9, American Gas Association,
1998.
16. Kegel, T. M., "Statistical Control of the Measure-
ment Process," 6th Pipeline Conference, Merida,
Yucatan, Mexico, Nov. 14-16, 2001.
17. Kegel, T. M., "Quality Control Program of the
CEESI Ventura Calibration Facility," FLOMEKO,
Gronigen, The Netherlands, May 2003.
t n e n o p m o C
d r a d n a t S
] % [ y t n i a t r e c n U ] % [ n o i t u b i r t n o C
e r u s s e r P
n o i t a c i f i c e p S e c n a m r o f r e P d e n i b m o C 4 1 1 . 0 0 8 . 2
y t i l i b a t S 4 0 2 . 0 5 9 . 8
t c e f f E e r u t a r e p m e T t n e i b m A 4 4 0 . 0 2 4 . 0
e r u s s e r P c i r t e m o r a B 1 1 0 . 0 3 0 . 0
n o i t i s i u q c A a t a D 3 9 0 . 0 5 8 . 1
e r u t a r e p m e T
n o i t a c i f i c e p S e c n a m r o f r e P d e n i b m o C 9 2 0 . 0 9 1 . 0
y t i l i b a t S 9 2 0 . 0 9 1 . 0
t c e f f E e r u t a r e p m e T t n e i b m A 6 0 0 . 0 1 0 . 0
e b o r P D T R 3 7 0 . 0 3 1 . 1
s t c e f f E r e f s n a r T t a e H 2 0 0 . 0 < 1 0 . 0
n o i t i s i u q c A a t a D 6 1 1 . 0 8 8 . 2
y t i l i b i s s e r p m o C e s a B
n o i t a u q E e t a t S 0 6 0 . 0 7 7 . 0
h p a r g o t a m o r h C 2 0 0 . 0 < 1 0 . 0
y t i l i b i s s e r p m o C g n i w o l F
n o i t a u q E e t a t S 0 6 0 . 0 7 7 . 0
h p a r g o t a m o r h C 0 9 1 . 0 6 7 . 7
e m u l o V g n i w o l F
e u l a V 7 A G A 0 8 5 . 0 7 2 . 2 7
Table 2: First Summary of Uncertainty Components
t n e n o p m o C
d r a d n a t S
] % [ y t n i a t r e c n U ] % [ n o i t u b i r t n o C
e r u s s e r P
n o i t a c i f i c e p S e c n a m r o f r e P d e n i b m o C 4 1 1 . 0 3 4 . 8
y t i l i b a t S 4 0 2 . 0 3 9 . 6 2
t c e f f E e r u t a r e p m e T t n e i b m A 4 4 0 . 0 5 2 . 1
e r u s s e r P c i r t e m o r a B 1 1 0 . 0 8 0 . 0
n o i t i s i u q c A a t a D 3 9 0 . 0 5 5 . 5
e r u t a r e p m e T
n o i t a c i f i c e p S e c n a m r o f r e P d e n i b m o C 9 2 0 . 0 6 5 . 0
y t i l i b a t S 9 2 0 . 0 6 5 . 0
t c e f f E e r u t a r e p m e T t n e i b m A 6 0 0 . 0 2 0 . 0
e b o r P D T R 3 7 0 . 0 0 4 . 3
s t c e f f E r e f s n a r T t a e H 2 0 0 . 0 < 1 0 . 0
n o i t i s i u q c A a t a D 6 1 1 . 0 5 6 . 8
y t i l i b i s s e r p m o C e s a B
n o i t a u q E e t a t S 0 6 0 . 0 3 3 . 2
h p a r g o t a m o r h C 2 0 0 . 0 < 1 0 . 0
y t i l i b i s s e r p m o C g n i w o l F
n o i t a u q E e t a t S 0 6 0 . 0 3 3 . 2
h p a r g o t a m o r h C 0 9 1 . 0 3 3 . 3 2
e m u l o V g n i w o l F
y r o t a r o b a L 5 2 1 . 0 0 1 . 0 1
y t i l i b i c u d o r p e R d n a y t i l i b a t a e p e R 0 0 1 . 0 6 4 . 6
Table 3: Second Summary of Uncertainty Components

S-ar putea să vă placă și