Iso 5168 - 1998

British Standard
A single copy of this

British Standard is licensed to
DANIEL EUROPE
on July 11, 2000
This is an uncontrolled copy.

Ensure use of the most current
version of this standard by
searching British Standards Online
at bsonline.techindex.co.uk
BRITISH STANDARD BS ISO TR
5168:1998
Measurement of fluid
flow — Evaluation of
uncertainties
Licensed Copy: DANIEL EUROPE, Daniel Europe Ltd, 11-Jul-00, Uncontrolled Copy. © BSI
ICS 17.120.01; 17.120.10

BS ISO TR 5168:1998
National foreword
This British Standard reproduces verbatim ISO TR 5168:1998 and implements

it as the UK national standard. It supersedes BS 5844:1980 which is
withdrawn.
The UK participation in its preparation was entrusted by Technical Committee
CPI/30, Measurement of fluid flow in closed conduits, to Subcommittee
CPI/30/9, General topics, which has the responsibility to:
— aid enquirers to understand the text;
— present to the responsible international/European committee any
enquiries on the interpretation, or proposals for change, and keep the UK
interests informed;
— monitor related international and European developments and
promulgate them in the UK.
A list of organizations represented on this subcommittee can be obtained on
request to its secretary.
Cross-references
The British Standards which implement international or European
publications referred to in this document may be found in the BSI Standards
Catalogue under the section entitled “International Standards Correspondence
Index”, or by using the “Find” facility of the BSI Standards Electronic
Catalogue.
A British Standard does not purport to include all the necessary provisions of
a contract. Users of British Standards are responsible for their correct
application.
Compliance with a British Standard does not of itself confer immunity
from legal obligations.
Summary of pages
This document comprises a front cover, an inside front cover, pages i and ii,
the ISO/TR title page, pages ii to iv, pages 1 to 64, an inside back cover
and a back cover.
This standard has been updated (see copyright date) and may have had
amendments incorporated. This will be indicated in the amendment table on
the inside front cover.
This British Standard, having Amendments issued since publication

been prepared under the
direction of the Sector Board Amd. No. Date Comments
for Materials and Chemicals,
was published under the
authority of the Standards
Board and comes
into effect on
15 June 1998
© BSI 07-1999
ISBN 0 580 27553 1

BS ISO TR 5168:1998
Contents
Page
National foreword Inside front cover
Foreword iv
Text of ISO/TR 5168 1
© BSI 07-1999 i
blank
ii
BS ISO TR 5168:1998
Contents
Page
Foreword iv
Introduction 1
1 Scope 1
2 Normative references 1
3 Definitions and symbols 2
4 General principles of measurement uncertainty analysis 6
5 Identification and classification of elemental measurement error
sources 12
6 Estimation and presentation of elemental uncertainties 15
7 Combination and propagation of uncertainties 19
8 Calculation of uncertainty 22
9 Presentation of results 24
Annex A (normative) Small sample methods 28
Annex B (normative) Outlier treatment 30
Annex C (informative) Examples of estimation of uncertainty in airflow
measurement 33
Annex D (informative) Examples of estimating uncertainty in open
channel flow measurement 48
Annex E (informative) Example of flowrate measurement in circular
pipes 56
Annex F (informative) Bibliography Inside back cover
Figure 1 — Measurement error 2
Figure 2 — Illustration of terms relating to errors and uncertainties 4
Figure 3 — Systematic error 4
Figure 4 — Uncertainty interval x ± U (see also Figure 2) 7
Figure 5 — Measurement error (systematic, random) and accuracy 8
Figure 6 — Random error 9
Figure 7 — Basic measurement calibration hierarchy 13
Figure 8 — Data acquisition system 14
Figure 9 — Calibration should compensate for test meter systematic error 18
Figure 10 — Measurement uncertainty interval (UADD); symmetrical
systematic uncertainty and additive combination 23
Figure 11 — Measurement uncertainty; nonsymmetrical systematic
uncertainty and additive combination 24
Figure 12 — Three after-measurement uncertainty interval comparisons 25
Figure B.1 — Outlier outside the range of acceptable data 31
Figure B.2 — Result of outlier tests 33
Figure C.1 — Schematic of sonic nozzle flowmeter installation upstream
of a turbine engine 34
Figure C.2 — Typical calibration hierarchy 35
Figure C.3 — Temperature measurement calibration hierarchy 39
Figure C.4 — Typical thermocouple channel 40
Figure C.5 — Flowrate through a sonic nozzle 47
Figure D.1 — Schematic of velocity-area method of discharge measurement
(midsection method) 50
Figure D.2 — Triangular profile weir 54
Table 1 — Calibration hierarchy error sources 14
Table 2 — Data acquisition error sources 14
Table 3 — Data reduction error sources 14
Table 4 — Uncertainty intervals defined by nonsymmetrical systematic
uncertainties 23
ii © BSI 07-1999
BS ISO TR 5168:1998
Page
Table 5 — Recommended corrective action if the predicted
before-measurement uncertainty is unacceptable 25
Table 6 — Elemental error sources 27
Table A.1 — Two-tailed Student’s t table for small sample
methods, < 30 degrees of freedom 30
Table B.1 — Rejection values for Grubbs’ method 31
Table B.2 — Results of Grubbs’ test 32
Table C.1 — Calibration hierarchy error sources 35
Table C.2 — Pressure transducer data acquisition error sources 36
Table C.3 — Pressure measurement data reduction error sources 37
Table C.4 — Temperature calibration hierarchy elemental errors 38

Table C.5 — Airflow measurement uncertainty sources 43
Table C.6 — Uncertainty comparisons of Examples 1 and 2 46
Table C.7 — Flowrate data 48
Table D.1 — Elemental uncertainties affecting uncertainty in discharge 53
Table D.2 — Error element values 55
Table E.1 — List of independent sources of error 57
Table E.2 — Nature of errors 58
Descriptors: Fluid flow, pipe flow, flow measurement, rules of calculation, error analysis.
© BSI 07-1999 iii

BS ISO TR 5168:1998
Foreword
ISO (the International Organization for Standardization) is a worldwide
federation of national standards bodies (ISO member bodies). The work of
preparing International Standards is normally carried out through ISO technical
committees. Each member body interested in a subject for which a technical
committee has been established has the right to be represented on that committee.
International organizations, governmental and non-governmental, in liaison with
ISO, also take part in the work. ISO collaborates closely with the International
Electrotechnical Commission (IEC) on all matters of electrotechnical
standardization.
The main task of technical committees is to prepare International Standards, but
in exceptional circumstances a technical committee may propose the publication of
a Technical Report of one of the following types:
— type 1, when the required support cannot be obtained for the publication of
an International Standard, despite repeated efforts;
— type 2, when the subject is still under technical development or where for any
other reason there is the future but not immediate possibility of an agreement
on an International Standard;
— type 3, when a technical committee has collected data of a different kind from
that which is normally published as an International Standard (“state of the
art”, for example).
Technical Reports of types 1 and 2 are subject to review within three years of
publication, to decide whether they can be transformed into International
Standards. Technical Reports of type 3 do not necessarily have to be reviewed until
the data they provide are considered to be no longer valid or useful.
ISO/TR 5168, which is a Technical Report of type 1, was prepared by Technical
Committee ISO/TC 30, Measurement of fluid flow in closed conduits,
Subcommittee SC 9, Uncertainties in flow measurement.
This document is being issued as a type 1 Technical Report because no consensus
could be reached between ISO TC 30/SC 9 and ISO/TAG 4, Metrology, concerning
the harmonization of this document with the Guide to the expression of uncertainty
in measurement, which is a basic document in the ISO/IEC Directives. A future
revision of this Technical Report will align it with the Guide.
This first edition as a Technical Report cancels and replaces the first edition as an
International Standard (ISO 5168:1978), which has been technically revised.
Annex A and Annex B form an integral part of this Technical Report. Annex C,
Annex D, Annex E and Annex F are for information only.
iv © BSI 07-1999
BS ISO TR 5168:1998
Introduction
One of the first International Standards to specifically address the subject of uncertainty in measurement
was ISO 5168, Measurement of fluid flow — Estimation of uncertainty of a flow-rate measurement,
published in 1978. The extensive use of ISO 5168 in practical applications identified many improvements
to its methods; these were incorporated into a draft revision of this International Standard, which in 1990
received an overwhelming vote in favour of its publication. However, this draft revision of ISO 5168 was
withheld from publication for a number of years since, despite lengthy discussions, no consensus could be
reached with the draft version of a document under development by a Working Group of ISO Technical
Advisory Group 4, Metrology (ISO TAG 4/WG 3). The TAG 4 document, Guide to the expression of
uncertainty in measurement (GUM), was published in late 1993 as a basic document in the ISO/IEC
Directives. At a meeting of the ISO Management Board in May 1995 it was decided to publish the revision
of ISO 5168, Measurement of fluid flow — Evaluation of uncertainties, as a Technical Report.
One of the major differences between ISO/TR 5168 and the GUM is in the definitions and terminology. In
addition, a substantial difference exists with respect to the concepts to be used to define practical
measurement processes. In this Technical Report a normal distribution of the measurement data is
assumed and Student’s t-factor is used to determine the uncertainty. The method used to propagate
elemental uncertainties to the overall uncertainty is essentially identical to that used in the GUM.
This document is published as a type 1 Technical Report instead of an International Standard because it
is not consistent with the GUM. A future revision of this Technical Report will align the two documents.
1 Scope
Whenever a measurement of flowrate (discharge) is made, the value obtained from the experimental data
is simply the best possible estimate of the true flowrate. In practice, the true flowrate may be slightly
greater or less than this value.
1.1 This Technical Report details step-by-step procedures for the evaluation of uncertainties in individual
flow measurements arising from both random and systematic error sources and for the propagation of
component uncertainties into the uncertainty of the test results. These procedures enable the following
processes to be carried out:
a) estimation of the accuracy of results derived from flowrate measurement;
b) selection of a proper measuring method and devices to achieve a required level of accuracy of flowrate
measurement;
c) comparison of the results of measurement;
d) identification of the sources of errors contributing to a total uncertainty;
e) refinement of the results of measurement as data accumulate.
NOTE It is assumed that the measurement process is carefully controlled and that all calibration corrections have been applied.
1.2 This Technical Report describes the calculations required in order to arrive at an estimate of the
interval within which the true value of the flowrate may be expected to lie. The principle of these
calculations is applicable to any flow measurement method, whether the flow is in an open channel or in a
closed conduit.
NOTE Although this Technical Report has been drafted taking mainly into account the sources of error due to the instrumentation,
it should be emphasized that the errors due to the flow itself (velocity distribution, turbulence, etc.) and to its effect on the method
and on the response of the instrument can be of great importance with certain methods of flow measurement (see 5.7). Where a
particular device or technique is used, some simplifications may be possible or special reference may have to be made to specific
sources of error not identified in this Technical Report. Therefore reference should be made to the “Uncertainty of measurement”
clause of the appropriate International Standard dealing with that device or technique.
2 Normative references
The following standards contain provisions which, through reference in this text, constitute provisions of
this Technical Report. At the time of publication, the editions indicated were valid. All standards are
subject to revision, and parties to agreements based on this Technical Report are encouraged to investigate
the possibility of applying the most recent editions of the standards listed below. Members of IEC and ISO
maintain registers of currently valid International Standards.
ISO 5725-1:1994, Accuracy (trueness and precision) of measurement methods and results — Part 1: General
principles and definitions.
© BSI 07-1999 1
BS ISO TR 5168:1998
ISO 5725-2:1994, Accuracy (trueness and precision) of measurement methods and results — Part 2: Basic
method for the determination of repeatability and reproducibility of a standard measurement method.
ISO 5725-3:1994, Accuracy (trueness and precision) of measurement methods and results —
Part 3: Intermediate measures of the precision of a standard measurement method.
ISO 5725-4:1994, Accuracy (trueness and precision) of measurement methods and results — Part 4: Basic
methods for the determination of the trueness of a standard measurement method.
ISO 5725-6:1994, Accuracy (trueness and precision) of measurement methods and results — Part 6: Use in
practice of accuracy values.
3 Definitions and symbols

For the purposes of this Technical Report, the following definitions and symbols apply.
3.1 Definitions
3.1.1
correction
value which must be added algebraically to the indicated value to obtain the corrected result. It is
numerically the same as a known error, but of opposite sign
3.1.2
coverage
percentage frequency at which an interval estimate of a parameter contains the true value. That is, in
repeated sampling when the uncertainty interval provides 95 % coverage for each sample, over the long
run the intervals will contain the true value 95 % of the time
3.1.3
error
result of a measurement minus the (conventional) true value of the measurement. See Figure 1
NOTE The known parts of an error of measurement may be compensated by applying appropriate corrections. The error of the
corrected result can be characterized by an uncertainty.
3.1.4
estimate
value calculated from a sample of data as a substitute for an unknown population parameter
for example, the experimental standard deviation (s) is the estimate which describes the population
standard deviation (B)
Figure 1 — Measurement error
2 © BSI 07-1999
BS ISO TR 5168:1998
3.1.5
fossilization
creation of a fixed systematic error from a live random error when only a single calibration is relevant in
the calibration process
3.1.6
influence [sensitivity] coefficient
uncertainty propagated to the result due to unit uncertainty of the measurement (see subclause 7.4)
3.1.7
observed value
value of a characteristic determined as the result of an observation or test
3.1.8
random error
see Figure 2 and subclause 4.2
3.1.9
random uncertainty
component of the uncertainty associated with the random error. See Figure 2
3.1.10
statistical quality control chart
chart on which limits are drawn and on which are plotted values of any statistic computed from successive
samples of a population
the statistics which are used (mean, range, percent defective, etc.) define the different kinds of control
charts
3.1.11
systematic error
see Figure 2 and Figure 3 and subclause 4.3
3.1.12
systematic uncertainty
component of the uncertainty associated with the systematic error. See Figure 2
3.1.13
taylor’s series
power series to calculate the value of a function at a point in the neighbourhood of some reference point
the series expresses the difference or differential between the new point and the reference point in terms
of the successive derivatives of the function. Its form is:
where f r(a) denotes the value of the rth derivative of f(x) at the reference point x = a. Commonly, if the
series converges, the remainder Rn, is made infinitesimal by selecting an arbitrary number of terms and
usually only the first term is used
3.1.14
uncertainty
1) Half the uncertainty interval, for a symmetrical uncertainty interval.
2) The positive and negative components of a nonsymmetrical uncertainty interval, denoted by U+ and
U– respectively.
3.1.15
uncertainty interval
estimate characterizing the range of values within which the true value of a measurand is expected to lie
© BSI 07-1999 3
BS ISO TR 5168:1998
3.1.16
welch-satterthwaite method
method for estimating degrees of freedom of the result when combining experimental standard deviations
with unequal degrees of freedom
3.1.17
working standard
standard, usually calibrated against a reference standard, which is used routinely to calibrate or check
material measures or measuring instruments
Figure 2 — Illustration of terms relating to errors and uncertainties
Figure 3 — Systematic error
4 © BSI 07-1999
BS ISO TR 5168:1998
3.2 Symbols
Symbol Meaning
B Systematic uncertainty of a symmetrical uncertainty interval.
Bij Elemental systematic uncertainty. The j subscript indicates the category, i.e.:
j = 1 calibration
= 2 data acquisition
= 3 data reduction
= 4 method
= 5 subjective or personal
The i subscript is the number assigned to a given elemental source of error. If i is more than
a single digit, a comma is used between i and j.
B+ , B– Positive and negative systematic uncertainties of a nonsymmetrical uncertainty interval.
overbar (–) Mean value (of a variable).

M Number of redundant instruments or tests.
N Sample size.
2
s Unbiased estimate of the variance, Ö2.
sij Estimate of the experimental standard deviation from one elemental source. The subscripts
are the same as the elemental systematic uncertainties in Bij.
sx s
Experimental standard deviation of the mean; equal to ---------
N
Spooled =
where
x i is the arithmetic mean of all xi at the jth datum point.
t95 Student’s statistical parameter at the 95 % confidence level. The degrees of freedom, v, of
the sample estimate of the experimental standard deviation are needed to obtain the t
values.
U+, U– Positive and negative uncertainties of a nonsymmetrical uncertainty interval.
UADD = B + t 95 s x
URSS =
© BSI 07-1999 5
BS ISO TR 5168:1998
x Arithmetic mean of the data values; xi.
xi Value of x at the ith datum point.
xij Value of xi at the jth datum point.
Y Arithmetic mean of the n measurements of the variable Y.

Yi A basic measurement.
¶ Systematic error, the fixed, or constant component of the total error, ¸.
¹ Difference between measurements.
¸ Total error.
¼ Random error.
Ûi Influence coefficient ·R/·Yi.
È Population mean.
Ö 2 Variance, the square of the standard deviation.
Subscripts
ADD Additive model.
RSS Root-sum-square model.
NOTE In ISO 5168:1978 and in many standards for flowrate measurement, e is used to indicate absolute uncertainty and E is used
for relative uncertainty. In this Technical Report U9 is used for relative uncertainty.
4 General principles of measurement uncertainty analysis

4.1 Nature of errors
All measurements have errors even after all known corrections and calibrations have been applied. The
errors may be positive or negative and may be of a variable magnitude. Many errors vary with time. Some
have very short periods while others vary daily, weekly, seasonally or yearly. Those which remain constant
or apparently constant during the test are called systematic errors. The actual errors are rarely known;
however, upper bounds on the errors can be estimated. The objective is to construct an uncertainty interval
(or sometimes referred to as range) within which the true value will lie with a stated probability.
Errors are the differences between the measurements and the true value which is always unknown. The
total measurement error, ¸, is divided into two components: ¶, a fixed systematic error and a random
error, ¼, as shown in Figure 2. In some cases, the true value may be arbitrarily defined as the value that
would be obtained by a specific metrology laboratory. Uncertainty is an estimate of the error which in most
cases would not be exceeded. There are three types of error to be considered:
a) random errors — see 4.2;
b) systematic errors — see 4.3;
c) spurious errors or mistakes (assumed to be identified and rejected prior to statistical analysis) —
see 4.4.
It is rarely possible to give an absolute upper limit to the value of the error. It is, therefore, more practicable
to give an interval within which the true value of the measured quantity can be expected to lie with a
suitably high probability. This “uncertainty interval” is shown as [ x – U, x + U ] in Figure 4 (the interval is
twice the calculated uncertainty).
Since measurement systems are subject to two types of errors, systematic and random, it follows that an
accurate measurement is one that has both small random and small systematic errors (see Figure 5).
6 © BSI 07-1999
BS ISO TR 5168:1998
4.2 Random error

Random errors are caused by numerous, small, independent influences which prevent a measurement
system from delivering the same reading when supplied with the same input value of the quantity being
measured. The data points deviate from the mean in accordance with the laws of chance, such that the
distribution usually approaches a normal distribution as the number of data points is increased. Random
errors are sometimes referred to as precision errors. The standard deviation (B) (see Figure 6) is used as a
measure of the random error, ¼. A large standard deviation means large scatter in the measurements. The
statistic (s) is calculated from a sample to estimate the standard deviation and is called the experimental
standard deviation.
. . . (1)
where
N is the number of measurements;
x is the average value of individual measurements x.
Figure 4 — Uncertainty interval x ± U (see also Figure 2)
© BSI 07-1999 7
BS ISO TR 5168:1998
Figure 5 — Measurement error (systematic, random) and accuracy
For the normal distribution, the interval x ± t 95 s / N will include the true mean, È, approximately 95 %
of the time. The random uncertainty of the mean is t 95 s / N . When the sample size is small, it is necessary
to use the Student’s t value at the 95 % level. For sample sizes equal to or greater than 30, two
experimental standard deviations (2s) are used as an estimate of the random uncertainty in an individual
measurement. This is explained in Annex A.
The random uncertainty can be reduced by making as many measurements as possible and using the
arithmetic mean value, since the standard deviation of the mean of N independent measurements is N
times smaller than the standard deviation of the measurements themselves.
Ö individual
Ö average = ----------------------------
- . . . (2)
N
and, analogously
s
s x = --------- . . . (3)
N
8 © BSI 07-1999
Licensed Copy: DANIEL EUROPE, Daniel Europe Ltd, 11-Jul-00, Uncontrolled Copy. © BSI BS ISO TR 5168:1998
Figure 6 — Random error

4.3 Systematic error
The second component of the total error is the systematic error, ¶. At each flow level this error is constant
for the duration of the test (Figure 1). In repeated measurements of a given sample, each measurement has
the same systematic error. The systematic error can be determined only when the measurements are
compared with the true value of the quantity measured and this is rarely possible. Systematic errors are
sometimes referred to as biases.
Every effort shall be made to identify and account for all significant systematic errors. These may arise
from (1) imperfect calibration corrections, (2) imperfect instrumentation installation, (3) imperfect data
reduction, and may include (4) method errors, and (5) human errors. As the true systematic error is never
known, an upper limit, B, is used in the uncertainty analysis.
In most cases, the systematic error, ¶, is equally likely to be plus or minus about the measurement. That
is, it is not known if the systematic error is positive or negative, and the systematic uncertainty reflects
this as ± B. The systematic uncertainty, B, is estimated as an upper limit of the systematic error, ¶.
4.4 Spurious errors
Spurious errors are errors, such as human mistakes or instrument malfunction, which invalidate a
measurement; for example, the transposing of numbers in recording data or the presence of pockets of air
in leads from a water line to a manometer. Such errors cannot be treated with statistical analysis and the
measurement should be discarded. Every effort should be made to eliminate spurious errors to properly
control the measurement process.
To ensure control, all measurements should be monitored with statistical quality control charts. Drifts,
trends and movements leading to out-of-control situations should be identified and investigated. Histories
of data from calibrations are required for effective control. It is assumed herein that these precautions are
observed and that the measurement process is under control; if not, the methods described are invalid.
After all obvious mistakes have been corrected or removed, there may remain a few observations which are
suspicious solely because of their magnitude.
For errors of this nature, the statistical outlier tests given in Annex B should be used. These tests assume
the observations are normally distributed. It is necessary to recalculate the experimental standard
deviation of the distribution of observations whenever a datum is discarded as a result of the outlier test.
It should also be emphasized that outliers should not be discarded unless there is an independent technical
reason for believing that spurious errors may exist: data should not lightly be thrown away.
4.5 Combining elemental uncertainties
The test objective, test duration and the number of calibrations related to the test affect the classification
of uncertainties into systematic and random components. Guidelines are presented in clause 6.
© BSI 07-1999 9
BS ISO TR 5168:1998
After all elemental error sources have been identified and classified as calibration, data acquisition, data
reduction, methodic and subjective error sources and elemental standard deviations and systematic
uncertainties estimated for each error source, a method for combining these elemental components into the
experimental standard deviation and systematic uncertainty of the measurement is needed. The
root-sum-square or quadrature combination is recommended.
. . . (4)
. . . (5)
4.6 Uncertainty of measurements

The measurement uncertainty analysis will be completed when:
a) the systematic uncertainties and standard deviations of the measure have been propagated to
uncertainty in the test result, keeping systematic and random components separate;
b) if small samples are involved, an estimate of the degrees of freedom of the experimental standard
deviation of the test result has been calculated from the Welch-Satterthwaite formula (see Annex A);
c) the random and systematic uncertainties are combined into a single number to express a reasonable
value for the overall uncertainty.
For simplicity of presentation, a single number, U, is needed to express a reasonable limit of error. The
single number, some combination of the systematic and random uncertainties, must have a simple
interpretation (e.g. the largest error reasonably expected), and be useful without complex explanation. For
example, the true value of the measurement is expected to lie within the interval
[ x – U, x + U ] . . . (6)
Since systematic uncertainties include those based on judgement and not on data, there is no way of
combining systematic and random uncertainties to produce a single uncertainty figure with a statistically
rigorous confidence level. However, since it is accepted that a single figure for the uncertainty of a
measurement is often required, two alternative methods of combination are permitted:
1) linear addition:
. . . (7)
2) root-sum-square combination:
. . . (8)
where B is the systematic uncertainty from equation (5) and sx is the experimental standard deviation of
the mean [equations (4) and (3)]. If large samples (N > 30) are used to calculate s, the value 2,0 may be used
for t95 for simplicity. If small samples (N < 30) are used to calculate s, the methods in Annex A are required.
There are two situations where it is possible to use a statistical confidence level for the uncertainty
interval:
a) if the systematic uncertainty is based on interlaboratory comparisons (see ISO 5725); and
b) if the systematic uncertainty is judged to be negligible compared to the random uncertainty. Here the
uncertainty interval is the test result which is at the 95 % confidence level.
Typically URSS is considered to have coverage of approximately 95 %, and UADD is considered to have
coverage between 95 % and 99 %.
10 © BSI 07-1999
BS ISO TR 5168:1998
4.7 Propagation of measurement uncertainties to test result uncertainties

If the test result is a function of several measurements, the experimental standard deviations and
systematic uncertainties of the measurements must be combined or propagated to the test result using
sensitivity factors, Û, that relate the measurement to the test result (see 7.4). Small sample methods are
given in Annex A.
In general, for m measurements, the experimental standard deviation and systematic uncertainty of the
test result are obtained as follows:
. . . (9)
. . . (10)
The overall uncertainty for the test result is formed in the same manner as described for the measurement
in 4.6.
4.8 Uncertainty analysis before and after measurement
Uncertainty analysis before measurement allows corrective action to be taken prior to the test to reduce
uncertainties when they are too large or when the difference to be detected in the test is the same size or
smaller than the predicted uncertainty. Uncertainty analysis before the test identifies the most
cost-effective corrective action and the most accurate measurement method.
The before-measurement uncertainty analysis is based on data and information that exists before the test,
such as calibration histories, previous tests with similar instrumentation, prior measurement uncertainty
analysis, expert opinions and, if necessary, special tests. With complex tests, there may be alternatives to
evaluate prior to the test such as different test designs, instrumentation arrangements, alternative
calculation procedures and concomitant variables. Corrective action resulting from this
before-measurement analysis may include:
a) improvements to instrumentation if the uncertainties are unacceptably high;
b) selection of a different measurement or calibration method;
c) repeated testing and/or increased sample sizes if the random uncertainties are unacceptably high. The
experimental standard deviation of the mean is reduced as the number of samples used to calculate the
mean is increased;
d) instead of repeated testing, the test duration may be extended in order to average the output
scatter (noise) of the flowmeter, resulting in a smaller random error per observation and hence a smaller
random uncertainty;
NOTE For example, ultrasonic and vortex shedding meters may have to be calibrated against a master meter, allowing longer
test times than allowed by a compact prover.
e) rotating flowmeters normally generate an output showing a periodic cycle superimposed on an
average meter factor. In this case the test duration shall be matched to an integer multiple of half or full
periodic intervals in order to obtain the shortest test times.
After-measurement analysis is based on the actual measurement data. It is required to establish the final
uncertainty. It is also used to confirm the before-measurement estimates and/or to identify data validity
problems. When redundant instrumentation or calculation methods are available, the individual
uncertainties should be compared for consistency with each other and with the before-measurement
uncertainty analysis. If the uncertainty intervals do not overlap, a problem is indicated. The
after-measurement random uncertainties should be compared with the before-measurement predictions.
© BSI 07-1999 11
BS ISO TR 5168:1998
5 Identification and classification of elemental measurement error sources

5.1 Summary of procedure
Make a complete list of every possible source of measurement error for all measurements that affect the
end test result. For convenience, group them by some or all of the following categories:
a) calibration,
b) data acquisition,
c) data reduction,
d) errors of method and
e) subjective or personal.
Within each category, there may be systematic and/or random error.
5.2 Systematic versus random
Systematic errors are those which remain constant in the process of measurement for a given value of
flowrate.
Typical examples of systematic errors of flowrate measurements are:
a) errors from a single flowmeter calibration;
b) errors of determination of the constants in the working formula of a measuring method;
c) errors due to truncating instead of rounding off the results of measurement.
Where the value and sign of a systematic error are known, it is assumed to be corrected (the correction
being equal in value and opposite in sign to the systematic error). Inaccuracy of the correction results in a
residual systematic uncertainty.
Random errors are those that produce variation (not predictable) in repeated measurements of the same
quantity.
Typical random errors associated with flowrate measurement are those caused by inaccurate reading of the
scale of a measuring instrument or by the scatter of the output signal of an instrument.
The effect of random errors on the random uncertainty may be reduced by averaging multiple results of the
same value of the quantity.
The preliminary decision to determine if a given elemental source contributes to systematic uncertainty,
random uncertainty or both, is made by adopting the following recommendation: the uncertainty of a
measurement should be put into one of two categories depending on how the uncertainty is derived. The
value of a random uncertainty is derived by a statistical analysis of repeated measurements, while that of
a systematic uncertainty is estimated by nonstatistical methods.
This recommendation avoids a complex decision and keeps the statistical estimates separate from the
judgement estimates as long as possible. The decision is preliminary and will be reviewed after
consideration of the defined measurement process.
5.3 Categorization of elemental error sources
Possible error sources can be divided arbitrarily into three to five categories:
a) calibration (see 5.4);
b) data acquisition (see 5.5);
c) data reduction (see 5.6);
d) method-related (see 5.7);
e) subjective or personal (see 5.8).
The size and complexity of the measurement uncertainty analysis may lead to the use of any or all of these
categories.
For example, metrological maintenance (calibration, verification, certification) of flowmeters, flowrate
measurements and processing of the data are done by different personnel. To control the possible sources
of error, it is advisable to relate them to the stages of preparation, measurement and processing of the data.
In such cases, it is advisable to classify error sources into:
a) calibration error sources (see 5.4);
12 © BSI 07-1999
BS ISO TR 5168:1998
b) errors of measurement or data acquisition error sources (see 5.5);

c) errors of processing the measurement data or data reduction error sources (see 5.6).
5.4 Calibration error sources
The major purpose of the calibration process is to determine systematic errors in order to eliminate them.
Thus, the calibration process exchanges the large systematic uncertainty of an uncalibrated or poorly
calibrated instrument for the smaller combination of the systematic uncertainty of the reference
instrument and the random uncertainty of the comparison. This exchange of uncertainties is fundamental
and is the basis of the notion that the uncertainty of the standard should be substantially less than that of
the test instrument.
Figure 7 shows a typical transducer calibration hierarchy. Each calibration in this hierarchy constitutes
an error source, with which is associated a pair of elemental uncertainties — the systematic uncertainty
and the experimental standard deviation of the process. It should be noted that, from one step to another,
these elemental uncertainties, as listed in Table 1, may be cumulative or independent. For example, B21
may include B11. The second digit of the subscript indicates the category, i.e. 1 indicates calibration.
Figure 7 — Basic measurement calibration hierarchy

5.5 Data acquisition error sources
Figure 8 illustrates some of the error sources associated with a typical pressure data acquisition system.
Data are acquired by measuring the electrical output resulting from pressure applied to a strain gauge type
pressure-measurement instrument. Other error sources, such as probe errors, installation effects and
environmental effects, also may be present. The effects of these error sources should be determined by
performing overall system calibrations, comparing known applied pressures with measured values.
However, should it not be possible to do this, then it is necessary to estimate each of the elemental
uncertainties and combine them to determine the overall uncertainty.
Some of the data acquisition error sources are listed in Table 2. Symbols for the elemental systematic
uncertainties and the experimental standard deviations and for the degrees of freedom are shown. Note
these elemental uncertainties are independent, not cumulative.
© BSI 07-1999 13
BS ISO TR 5168:1998
Table 1 — Calibration hierarchy error sources

Experimental Degrees
Systematic
Calibration standard of
uncertainty
deviation freedom
SL — ILS B11 s11 v11

ILS — TS B21 s21 v21
TS — WS B31 s31 v31
WS — MI B41 s41 v41
Table 2 — Data acquisition error sources

Systematic
Error source standard of
uncertainty
deviation freedom
Excitation voltage B12 s12 v12

Signal conditioning B22 s22 v22
Recording device B32 s32 v32
Pressure transducer B42 s42 v42
Probe errors B52 s52 v52
Environmental effects B62 s62 v62
Spatial averaging B72 s72 v72
Figure 8 — Data acquisition system

5.6 Data reduction error sources
Computations on raw data produce output in the required engineering units. Typical errors in this process
stem from curve fits and computational resolution. Uncertainties associated with these error sources are
often negligible.
Symbols for the data reduction error sources are listed in Table 3.
Table 3 — Data reduction error sources
Systematic
Calibration standard of
uncertainty
deviation freedom
Curve fit B13 s13 v13
Computational B23 s23 v23
resolution
14 © BSI 07-1999
BS ISO TR 5168:1998
5.7 Method error sources

Errors of method are those associated with a particular measurement procedure (principles of use of
instruments) and also with the uncertainty of constants used in calculations.
Some examples are errors from indirect methods of flowrate measurement associated with physical
inaccuracy of the relationship between the measured quantity and flowrate, or with inaccuracy of the
constants in the relationship. These inaccuracies may be due, for instance, to the fact that the flow
conditions prevailing during the measurement are not identical to the conditions in which the calibration
has been carried out or for which a standardized discharge coefficient has been established. In certain
methods of flow measurement (differential pressure devices for instance), these sources of error arising
from the flow conditions are covered by the uncertainty associated with the discharge coefficient, as far as
the installation conditions prescribed in the standard are satisfied; if they are not, that standard does not
apply. In other methods (velocity-area method for instance), the uncertainty arising from the flow
conditions is identified as a component of the total uncertainty; it shall be evaluated by the user in each
case and combined with the other elemental uncertainties.
5.8 Subjective error sources
Subjective error sources are caused by personal characteristics of the operators who calibrate flowmeters,
perform measurements and process the data. These can include reading errors and miscalculations.
6 Estimation and presentation of elemental uncertainties

Obtain an estimate of each elemental uncertainty. If the data is available to estimate the experimental
standard deviation, classify the uncertainty as a random uncertainty. Otherwise, classify it as a systematic
uncertainty.
Review the test objective, test duration and number of calibrations that will affect the test result. Make the
final classification of elemental uncertainties for each measurement. If an error increases the scatter in the
measurement result in the defined test, it is a random error; otherwise, it is a systematic error.
6.2 Calculation of experimental standard deviation
There are many ways to calculate the experimental standard deviation.
a) If the parameter to be measured can be held constant, a number of repeated measurements can be
used to evaluate equation (1), repeated here:
. . . (11)
b) If there are M redundant instruments or M redundant measurements and the parameter to be

measured can be held constant to take N repeat readings, the following pooled estimate of the
experimental standard deviation for individual readings can be used:
. . . (12)
NOTE Here ij is used differently from elsewhere in the text.
© BSI 07-1999 15
BS ISO TR 5168:1998
For the experimental standard deviation of the average value of the parameter
. . . (13)
c) If a pair of instruments (providing measurements x1i and x2i which have the same experimental
standard deviation are used to estimate a parameter that is not constant with time, the difference
between the readings, %, may be used to estimate the experimental standard deviation of the individual
instruments as follows:
. . . (14)
where
If the degrees of freedom are less than 30, the small sample methods shown in Annex A are required.
6.3 Estimation of systematic uncertainty
In spite of applying all known corrections to overcome imperfections in calibration, data acquisition and
data reduction processes, some systematic errors will probably remain. To determine the exact systematic
error in a measurement, it would be necessary to compare the true value and the measurements. However,
as the true value is unknown, it is necessary to carry out special tests or utilize existing data that will
provide systematic uncertainty information. The following examples are given, in order of preference.
a) Interlaboratory or interfacility tests make it possible to obtain the distribution of systematic errors
between facilities (see ISO 5725).
b) Comparisons of standards with instruments in the actual test environment may be used.
c) Comparison of independent measurements that depend on different principles can provide systematic
uncertainty information. For example, in a gas turbine test, airflow can be measured with (1) an
orifice, (2) a bellmouth nozzle, (3) compressor speedflow rig data, (4) turbine flow parameters and (5) jet
nozzle calibrations.
d) When it is known that a systematic error results from a particular cause, calibrations may be
performed allowing the cause to vary through its complete range to determine the range of systematic
error.
e) If there is no source of data on which to estimate the systematic uncertainty, the estimate must be
based on judgement. An estimate of an upper limit of the systematic error is needed. Instrumentation
manufacturers’ reports and other references may provide information. It is important to distinguish
between the “estimate” of an upper limit on systematic error obtained by this method and the more
reliable estimate of a random uncertainty arrived at by analysing data. There is a general tendency to
underestimate systematic uncertainties when a subjective approach is used, partly through human
optimism and partly through the possibility of overlooking the existence of some sources of systematic
error. Great care is therefore necessary when quoting systematic uncertainties.
f) If the mean of results from redundant instruments or measurements differs by more than has been
predicted by individual uncertainties, then a source of systematic uncertainty has been overlooked.
16 © BSI 07-1999
BS ISO TR 5168:1998
Sometimes the physics of the measurement system provide knowledge of the sign but not the magnitude of
the systematic error. For example, hot thermocouples radiate and conduct thermal energy from the sensor
to indicate lower temperatures. The systematic uncertainties interval in this case is nonsymmetrical,
i.e. not of the form ± B. It is of the form B+ for the positive and B– for the negative uncertainty. Thus, typical
systematic uncertainties associated with a radiating thermocouple could be:
B+ = 0°
B– = – 10°
For elemental uncertainties, the interval from B+ to B– shall include zero.

6.4 Final uncertainty classification based on the defined measurement process
Uncertainty statements must be related to a well-defined measurement process. The final classification of
uncertainties into systematic and random depends on the definition of the measurement process. Some of
these considerations are:
a) long versus short-term testing (see 6.4.1);
b) comparative versus absolute testing (see 6.4.2);
c) averaging to reduce random error (see 6.4.3).
6.4.1 Long versus short-term testing
The calibration histories accumulated before, during or after the testing period may influence the
uncertainty analysis.
a) When the instrumentation is calibrated only once, all the calibration uncertainty is frozen into
systematic uncertainty. The error in the calibration correction is a constant and cannot increase the
scatter in a test result. Thus, the calibration uncertainty, made up in general of systematic and fossilized
random uncertainties, is considered to be all systematic uncertainties in this case.
b) If the test period is long enough that instrumentation may be calibrated several times and/or several
test stands are involved, the random error in the calibration hierarchy (see 5.4) should be treated as
contributing to the overall experimental standard deviation. The experimental standard deviations may
be derived from calibration data.
6.4.2 Comparative versus absolute testing
The objective of a comparative test is to determine, with the smallest measurement uncertainty possible,
the net effect of a change. The first test is run with the standard or baseline configuration. The second test
is run with the change. The difference between the results of these tests is an indication of the effect of the
change. As long as only the difference or net effect between the two tests is considered, all systematic
errors, being fixed, will cancel out. The measurement uncertainty will be composed of random errors only.
All errors in a comparative test arise from random errors in data acquisition and data reduction.
Systematic errors are effectively zero. Since calibration random errors have been considered systematic
errors, they also are effectively zero.
The test result is the difference in flowrate between two test results, r1 and r2.
%r = r1 – r2 . . . (15)
and
. . . (16)
where sr is the experimental standard deviation of the random error of the first test, s¹r is the
1
root-sum-square of the experimental standard deviations from data acquisition and data reduction, and sr
2
is assumed to equal sr .
1
© BSI 07-1999 17
BS ISO TR 5168:1998
6.4.3 Averaging to reduce random error

Averaging test results is often used to improve the random uncertainty. Careful consideration should be
given to designing the test series to average as many causes of variation as possible within cost constraints.
The design should be tailored to the specific situation. For example, if experience indicates time-to-time
and rig-to-rig variations are significant, a design that averages multiple test measurement results on one
rig on one day may produce optimistic random uncertainty compared to testing several rigs, each mounted
several times, over a period of weeks. The list of possibilities may include the above plus test stand-to-test
stand, instrument-to-instrument, mount-to-mount and environmental, power supply and test crew
variation. Historic data is invaluable for studying these effects,1) If the pretest uncertainty analysis
identifies unacceptably large elemental uncertainties, special tests to measure the effects should be
considered.
6.5 Example of uncertainty classification: a calibration constant
The final classification of elemental uncertainty depends on the defined measurement process. To
illustrate, assume a test meter is to be compared or calibrated with a master meter at one flow level. The
objective is to determine a correction, called a calibration constant, that will be added to the test meter
observations when it is installed for test. This calibration constant correction will, over a limited time
period, make the test meter “read like” the master meter. During the calibration, the master meter is used
to set the flow level, as it is normally more accurate than the test meter. To reduce the calibration random
uncertainty, N = 13 comparisons will be made and averaged. If the data were plotted, the data might look
like Figure 9.
Figure 9 — Calibration should compensate for test meter systematic error
If the master meter systematic uncertainty derived from its own calibration is judged to be no larger than
BM, what will the test meter uncertainty be after calibration?
Define %i = Master meter readingi – Test meter readingi.
Calibration constant K equals the average
. . . (17)
1)
A statistical technique, analysis of variance (ANOVA) is useful for partitioning total variance by cause.
18 © BSI 07-1999
BS ISO TR 5168:1998
The experimental standard deviation of the calibration constant K is:
. . . (18)
The test meter is later installed in a test stand. Each observation made on the test meter is corrected by
adding K. By this process, the error in K from the calibration process is propagated to the corrected data
from the test stand.
If the defined measurement process is short, involving a single calibration, K is constant and this error is
a constant or systematic error. The uncertainty includes the systematic uncertainty in the master meter
plus the random uncertainty in the calibration process. The random uncertainty is fossilized into
systematic uncertainty. The fossilization is indicated by an asterisk. The systematic uncertainty may be
estimated:
. . . (19)
where
BM is the systematic uncertainty of the master meter;
t95 = 2,179 for 12 degrees of freedom (see Annex A).
This calibration systematic uncertainty should be combined with systematic uncertainties arising from
other sources to obtain the systematic uncertainty of the measurement. There may also be random
uncertainties arising from these other sources.
If the uncalibrated test meter had a systematic uncertainty judged to be BT, the calibration process
improved the test accuracy if BK is less than BT. Note that the calibration process does not change the test
meter random uncertainty, which is included in the data acquisition random uncertainty. However, the
test meter random uncertainty contributes to the calibration random uncertainty. This contribution is
reduced by averaging the calibration data.
If the test process is long and involves several calibrations, the calibration error contributes both
systematic uncertainty (BM) and random uncertainty ( t 95s K ) to the final test result.
If the test process is comparative (the difference between two tests with a single calibration), the calibration
error is all systematic error and cancels out when one result is subtracted from the other.
7 Combination and propagation of uncertainties

For each measurement, combine separately the elemental systematic uncertainties and the elemental
experimental standard deviations by the root-sum-square method. Propagate the measurement systematic
uncertainty and the experimental standard deviation separately all the way to the final test result, either
by sensitivity coefficients or by finitely incrementing the data reduction program. Work consistently in
either absolute units or percentages.
7.2 Combining elemental experimental standard deviations
The experimental standard deviation (s) of the measurement is the root-sum-square of the elemental
experimental standard deviations from all sources, that is:
. . . (20)
where j defines the category, such as (1) calibration, (2) data acquisition, (3) data reduction, (4) errors of
method and (5) subjective or personal, and i defines the sources within the categories.
© BSI 07-1999 19
BS ISO TR 5168:1998
For example, the experimental standard deviation for the calibration process in Table 1 is:
. . . (21)
The measurement experimental standard deviation is the root-sum-square of all the elemental
experimental standard deviations in the measurement system:
. . . (22)
7.3 Combining elemental systematic uncertainties

If there were only a few sources of elemental systematic uncertainties, it might be reasonable to add them
together to obtain the overall systematic uncertainties. For example, if there were three sources, the
probability that they would all be plus (or minus) would be one-half raised to the third power, or one-eighth.
However, the probability that all three will have the same sign and be at the limit of the systematic
uncertainties is extremely small. In actual practice, most measurements will have ten, twenty or more
sources of systematic error. The probability that they would all be plus (or minus) and be at their limit is
close to zero, and therefore it is more appropriate to combine them by root-sum-square.
If a measurement uncertainty analysis identifies four or less sources of systematic uncertainty, there
should be some concern that some sources have been overlooked. The analysis should be random and expert
help should be recruited to examine the calibration hierarchy, the data acquisition process and the data
reduction procedure for additional sources.
Therefore, the systematic uncertainty will be used herein as the root-sum-square of the elemental
systematic uncertainties from all sources.
. . . (23)
For example, the systematic uncertainty for the calibration hierarchy (Table 1) is
. . . (24)
The systematic uncertainty for the basic measurement process is
. . . (25)
If any of the elemental systematic uncertainties are nonsymmetrical, separate root-sum-squares are used
to obtain B+ and B–. For example, assume B21 + , B – , B + and B – are available. Then
21 23 23
. . . (26)
. . . (27)
20 © BSI 07-1999
BS ISO TR 5168:1998
7.4 Propagation of measurement uncertainties

Fluid flow parameters are rarely measured directly; usually more basic quantities such as temperature and
pressure are measured, and the fluid flow parameter is calculated as a function of the measurements.
Uncertainty of the measurements is propagated to the parameter through the function. The effect of the
propagation may be approximated with Taylor’s series methods. It is convenient to introduce the concept
of the sensitivity of a result to a measured quantity as the uncertainty propagated to the result due to unit
uncertainty of the measurement. The “sensitivity coefficient” (also known as “influence coefficient”) of each
subsidiary quantity is most easily obtained in one of two ways.

a) Analytically
Where there is a known mathematical relationship between the result, R, and subsidiary quantities Yi,
Y2, . . . YK, the dimensional sensitivity coefficient Ûi of the result R to the quantity Yi is obtained by
partial differentiation. Thus, if R = f(Y1, Y2, . . . YK,), then
. . . (28)
Analogously, the relative (nondimensional) sensitivity coefficient, Û i9 , is
. . . (29)
In this form, the sensitivity is expressed as “percent/percent”. That is, Û i9 is the percentage change in R
brought about by a 1% change in Yi . This is the form used if the uncertainties to be combined are
expressed as percentages of their associated variables rather than absolute values.
b) Numerically
Where no mathematical relationship is available or when differentiation is difficult, finite increments
may be used to evaluate Ûi. This is a convenient method with computer calculations.
Here Ûi is given by
. . . (30)
The result is calculated using Yi to obtain R, and then recalculated using (Yi + ¹Yi) to obtain (R +¹R).
The value of ¹Yi used should be as small as practicable.
If the experimental standard deviations of the measurement are small and the variables independent,
the experimental standard deviation of the result R is given by:
. . . (31)
and the systematic uncertainty by:
. . . (32)
Care should be taken to ensure that the variables are independent. With complex parameters, the same
measurement may be used more than once in the formula. This may increase or decrease the uncertainty
depending on whether the sign of the measurement is the same or opposite. If the Taylor’s series relates
the most elementary measurements to the ultimate parameter or result, these “linked” relationships will
be properly accounted for.
© BSI 07-1999 21
BS ISO TR 5168:1998
This effect can be covered by calculating a modified Û by simultaneous perturbation of all the inputs likely
to be affected, thus:
Ûlink = (change in output R due to a change in linked parameter which affects all inputs, Yi,
simultaneously)
An example of this is a change in barometric pressure, which affects all pressure inputs simultaneously in
a “gauge-pressure” system. Another example is the use of a common working standard to calibrate all the
pressure transducers.
The product of the sensitivity coefficient and the experimental standard deviation of the linked parameters
can then be combined with independent ones, thus:
. . . (33)
In equation (33), the product of one linked parameter is combined with the products of the remaining
independent parameters. The products of other linked parameters may be combined in a similar way.
Examples of the propagation of measurement uncertainties to a parameter can be found in Annex C.
8 Calculation of uncertainty
Select the additive or the root-sum-square model of combination and combine the systematic and random
uncertainties of the test result to obtain the overall uncertainty. The test result plus and minus the
uncertainty is the uncertainty interval that should contain the true value with high probability.
8.2 Uncertainty intervals
For simplicity of presentation, a single number (some combination of systematic and random uncertainties)
is needed to express a reasonable limit for error. The single number should have a simple interpretation
(e.g., the largest error reasonably expected) and be useful without complex explanation. It is usually
impossible to define a single rigorous statistic because the systematic uncertainty is based on judgement
which has unknown characteristics.2) This function is a hybrid combination of an estimated quantity based
on judgement (systematic uncertainty) and a statistic (random uncertainty). If both numbers were
statistics, a confidence interval would be recommended. 95 % or 99 % confidence levels would be available
at the discretion of the analyst. Although rigorous statistical confidence levels are not available, two
uncertainty intervals, with an associated coverage approximately analogous to 95 % and 99 % confidence
levels, are recommended.
8.3 Symmetrical intervals
Uncertainty (Figure 10) for the symmetrical systematic uncertainty case is centred about the
measurement, and the uncertainty interval is defined as:
R – U, R + U
where
. . . (34)
. . . (35)
If the experimental standard deviation is based on small samples, the methods in Annex A may be used to
determine a value of Student’s t95. For large samples (> 30), 2 may be substituted for t95 in equations (34)
and (35).
2)
If information exists to justify the assumption that the systematic uncertainties have a random distribution, a rigorous
statistic can be defined.
22 © BSI 07-1999
BS ISO TR 5168:1998
If the test result is an average ( R ) based on sample size N, instead of a single value (R), s/ N should be
substituted for s.
The uncertainty selected [equation (34) or (35)] should be provided in the presentation; the components
(systematic uncertainty, random uncertainty, degree of freedom) should be available in an appendix or in
supporting documentation. These three components may be required to substantiate and explain the
uncertainty value, to provide a sound technical base for improved measurements, and to propagate the
uncertainty from measured parameters to fluid flow parameters and from fluid flow parameters to other
more complex performance parameters (e.g. engine, pump or fan performance and power station or
chemical plant efficiency).
8.4 Nonsymmetrical interval
If there is a nonsymmetrical systematic uncertainty interval, the uncertainty (U) is no longer symmetrical
about the measurement. The interval is defined by the positive systematic uncertainty (B+) and the
negative systematic uncertainty (B–)(see 7.3).
Figure 11 shows the uncertainty (U) for nonsymmetrical systematic uncertainties. (See Table 4.)
Figure 10 — Measurement uncertainty interval (UADD); symmetrical systematic uncertainty

and additive combination
Table 4 — Uncertainty intervals defined by nonsymmetrical systematic uncertainties
B– B+ – U–
t 95 s x U ADD U+ RSS U+
ADD RSS
0K + 10 K 2K –2K + 12 K –2K + 10,2 K
– 3 kg + 13 kg 4 kg – 7 kg + 17 kg – 5 kg + 13,6 kg
0 Pa + 7 Pa 2 Pa – 2 Pa + 9 Pa – 2 Pa + 7,3 Pa
–8K 0K 2K – 10 K +2K – 8,2 K + 2,0 K
© BSI 07-1999 23
BS ISO TR 5168:1998
. . . (36)
. . . (37)
Figure 11 — Measurement uncertainty; nonsymmetrical systematic uncertainty and

additive combination
9 Presentation of results
9.1 Summary of requirement
The summary report should contain the nominal level of the test result, the systematic uncertainty, the
experimental standard deviation, the degrees of freedom and the overall uncertainty. The equation used to
calculate uncertainty, UADD or URSS should be stated. The summary should reference a table of the
elemental uncertainties considered and included in the uncertainty.
9.2 Before-measurement analysis and corrective action
Uncertainty is a function of the measurement process. It provides an estimate of the largest error that may
reasonably be expected for that measurement process. Errors larger than the uncertainty should rarely
occur. If the difference to be detected in an experiment is of the same size or smaller than the projected
uncertainty, corrective action should be taken to reduce the uncertainty. Therefore, it is recommended that
an uncertainty analysis always be done before the test or experiment. The recommended corrective action
depends on whether the systematic or the random uncertainty is too large as shown in Table 5.
24 © BSI 07-1999
BS ISO TR 5168:1998
Table 5 — Recommended corrective action if the predicted

before-measurement uncertainty is unacceptable
Systematic uncertainty too large: Random uncertainty too large:
• Improve calibration • Larger test sample

• Independent calibrations for redundant meters • More precise instrumentation
• Concomitant variable • Redundant instrumentation
• In-situ calibration • Data smoothing
— Moving average
— Filter
— Regression
• Improve design of experiment
9.3 After-measurement analysis and data validity

After-measurement analysis is required to confirm the before-measurement estimates or to identify data
validity problems. Comparison of measurement test results with the before-measurement analysis is an
excellent data validity check. The random uncertainty of the repeated points or redundant instruments
should not be significantly larger than the before-measurement estimates. When redundant
instrumentation or calculation methods are available, the individual uncertainty intervals should be
compared for consistency with each other and with the before-measurement uncertainty analysis.
Three cases are illustrated in Figure 12.
Figure 12 — Three after-measurement uncertainty interval comparisons

When there is no overlap between uncertainty intervals, as in Case I, a problem exists. The true value
cannot be contained within both intervals. That is, there should be a very low probability that the true
value lies outside any of the uncertainty intervals. Either the uncertainty analysis is wrong or a data
validity problem exists. Investigation to identify bad readings, overlooked sources of systematic error, etc.,
is necessary to resolve this discrepancy. Redundant and dissimilar instrumentation should be compared.
Partial overlap of the uncertainty intervals, as in Case II, also signals that a problem may exist. The
magnitude of the problem depends on the amount of overlap. The only situation when one can be confident
that the data is valid and the uncertainty analysis is correct is Case III, when the uncertainty intervals
attain maximum overlap.
© BSI 07-1999 25
BS ISO TR 5168:1998
9.4 Summary for reporting error

The definition of the components, systematic uncertainty, experimental standard deviation and the overall
uncertainty (U) suggests a summary format for reporting measurement uncertainty. The format will
describe the components of uncertainty, which are necessary to estimate further propagation of the
uncertainties, and a single value (U) which may be described as the largest error expected from the
combined uncertainties. Additional information (degrees of freedom for the estimate of s) is required to use
the experimental standard deviation if small samples were used to calculate s. These summary numbers
provide the information necessary to accept or reject the measurement uncertainty. The reporting format
should contain:
a) s, the experimental standard deviation, calculated from data;
b) for small samples, v, the degrees of freedom associated with the experimental standard deviation (s).
The degrees of freedom for small samples (less than 30) is obtained from the Welch-Satterthwaite
procedure illustrated in Annex A.
c) B, the systematic uncertainty of the measurement process or B– and B+ if the systematic uncertainty
is nonsymmetrical;
d) the uncertainty formula , the uncertainty interval within
which the true value is expected to lie. If the systematic uncertainty is nonsymmetrical,
and . No more than two
significant places should be reported. For small samples see Annex A.

The model components, s, v, B and U, are required to report the uncertainty of any measurement process.
The first three components, s, v, and B, are necessary: (1) to indicate corrective action if the uncertainty is
unacceptably large before the test, (2) to propagate the uncertainty to more complex parameters, and (3) to
substantiate the overall uncertainty.
9.5 Reporting uncertainty: Table of elemental sources
To support the measurement uncertainty summary, a table detailing the elemental error sources is needed
for several purposes. If the projected uncertainty is deemed excessive for the purpose of the test, corrective
action should be taken to reduce this uncertainty. Further, if the uncertainty quoted in the summary
appears to be optimistically small, the list of sources considered should be reviewed to identify missing
sources. For this reason, it is important to list all sources considered, even if negligible.
Table 6 gives the format for such a table; one line should be completed for each elemental error source. Note
that all uncertainties in this table have been propagated from the basic measurement to the end result.
26 © BSI 07-1999
BS ISO TR 5168:1998
Table 6 — Elemental error sources

Experimental Source of
Measurement Degrees of Systematic
Subscript Source standard systematic
ij value freedom uncertainty
deviation vij Bij uncertainty
sij
11
21
31
—
—
—
12
22
32
42
—
—
—
13
23
33
—
—
—
Measurement
Results v t95
value
© BSI 07-1999 27
BS ISO TR 5168:1998
Annex A (normative)
Small sample methods
A.1 Student’s t
When the experimental standard deviation is based on small samples (N < 30), uncertainty is defined as:
. . . (A.1)
. . . (A.2)
For these small samples, the interval will contain the true unknown
average, È, 95 % of the time. If the systematic uncertainty is negligible, this statistical confidence interval
is the uncertainty interval. t95 is the 95th percentile point for the two-tailed Student’s t-distribution. For
small samples, t will be large, and for larger samples t will be smaller, approaching 1,96 as a lower limit.
The t-value is a function of the number of degrees of freedom (v) used in calculating s. Since 30 degrees of
freedom (v) yield a t of 2,05 and infinite degrees of freedom yield a t of 1,96, an arbitrary selection of t = 2
is used for simplicity for values of v from 30 to infinity (see A.1).
A.2 Number of degrees of freedom for small samples
In a sample, the number of degrees of freedom (v) is equal to the sample size, N. When a statistic is
calculated from the sample, the degrees of freedom associated with the statistic is reduced by 1 for every
estimated parameter used in calculating the statistic. For example, from a sample of size N, x is calculated
and has N degrees of freedom, and the experimental standard deviation, s, is calculated using equation (1)
(see subclause 4.2), and has N–1 degrees of freedom because x is used to calculate s. In calculating other
statistics, more than one degree of freedom may be lost. For example, in calculating the experimental
standard deviation of a curve fit, the number of degrees of freedom is equal to N–k where k is the number
of estimated coefficients for the polynomial fitted to the data.
When all random uncertainties have large sample sizes (i.e. vij > 30) the calculation of number of degrees
of freedom (v) is unnecessary and 2 is substituted for t95. However, for small samples, when combining
experimental standard deviations by the root-sum-square method [see equation (21) in subclause 7.2 for
example], the number of degrees of freedom (v) associated with the combined experimental standard
deviations is calculated using the Welch-Satterthwaite formula (A.3).
For example: the number of degrees of freedom for the calibration experimental standard deviation (s1)
given by equation (21), is:
. . . (A.3)
where vi1 is the number of degrees of freedom of each elemental experimental standard deviation in the
calibration process.
28 © BSI 07-1999
BS ISO TR 5168:1998
The number of degrees of freedom for the measurement experimental standard deviation (s), as given by
equation (22), is:
. . . (A.4)
If the test result is an average, x , based on a sample of size N,
. . . (A.5)
As N is a known constant, the number of degrees of freedom of s x is the same as s, i.e.

. . . (A.6)
A.3 Propagating the degrees of freedom

The Student’s t value from A.1 to be used in calculating the uncertainty of the test result [equation (A.1)
or (A.2)] is based on vr, the number of degrees of freedom of sr. If the number of degrees of freedom of any
measurement standard deviation is less than 30, the number of degrees of freedom of the result also may
be less than 30. In such cases, the following small sample method may be used to determine vr. This is
defined for the absolute experimental standard deviation according to the Welch-Satterthwaite formula by:
. . . (A.7)
and for the relative experimental standard deviation by:
. . . (A.8)
where
and the number of degrees of freedom on the experimental standard deviation (sPi) of the independent
measurements is usually given by:
. . . (A.9)
NOTE The number of degrees of freedom for the relative and absolute experimental standard deviations are identical.
Welch-Satterthwaite degrees of freedom may contain fractional, decimal parts. The fractions should be
dropped or truncated, as rounding down is conservative with Student’s t, e.g. v = 13,6 should be treated as
v = 13,0.
© BSI 07-1999 29
BS ISO TR 5168:1998
Table A.1 — Two-tailed Student’s t table for small

sample methods, < 30 degrees of freedom
Number of Number of
degrees of t95 degrees of t95
freedom freedom
1 12,706 16 2,120
2 4,303 17 2,110
3 3,182 18 2,101
4 2,776 19 2,093
5 2,571 20 2,086
6 2,447 21 2,080
7 2,365 22 2,074
8 2,306 23 2,069
9 2,262 24 2,064
10 2,228 25 2,060
11 2,201 26 2,056
12 2,179 27 2,052
13 2,160 28 2,048
14 2,145 29 2,045
15 2,131 Â 1,96
Annex B (normative)
Outlier treatment
B.1 General
All measurement systems may produce spurious data points. These points may be caused by temporary or
intermittent malfunctions of the measurement system or they may represent actual variations in the
measurement. Errors of this type should not be included as part of the uncertainty of the measurement.
Such points are meaningless as test data. They should be discarded. Figure B.1 shows a spurious datum
point, called an outlier.
All data should be inspected for spurious data points as a continuing check on the measurement process.
Points should be rejected based on engineering analysis of instrumentation, fluid mechanics, flow profiles
and past history with similar data. To ease the burden of scanning large masses of data, computerized
routines are available to scan steady-state data and flag suspected outliers. The flagged points should then
be subjected to an engineering analysis.
The effect of these outliers is to increase the random uncertainty of the system. A test is needed to
determine if a particular point from a sample is an outlier. The test should consider two types of error in
detecting outliers:
1) rejecting a good datum point;
2) not rejecting a bad datum point.
The probability of rejecting a good point is usually set at 5 %. This means that the odds of rejecting a good
point are 20 to 1 (or less). The odds will be increased by setting the probability of 1) lower. However, this
practice decreases the probability of rejecting bad data points. The lower probability of rejecting a good
point will require that the rejected points be further from the calculated mean, and fewer bad data points
will thus be identified. For large sample sizes, e.g. containing several hundred measurements, almost all
bad data points can be identified. For small samples (five or ten measurements), bad data points are hard
to identify.
One test in common usage for determining whether spurious data are outliers is Grubbs’ method.
30 © BSI 07-1999
Figure B.1 — Outlier outside the range of acceptable data

B.2 Grubbs’ method
Consider a sample (xi) of N measurements. The mean ( x ) and an experimental standard deviation (s) are
calculated by equation (1). Suppose that (xj), the jth observation, is the suspected outlier; the absolute
statistic, Tn, calculated is then:
. . . (B.1)
Using Table B.1, a value of Tn is obtained for the sample size (N) at the 5 % significance level (P). This
limits the probability of rejecting a good point to 5 %. (The probability of not rejecting a bad datum point is
not fixed. It will vary as a function of sample size.)
The test for the outlier is to compare the calculated Tn with the value for Tn given in Table B.1.
If Tn calculated is larger than or equal to Tn as given in Table B.1, we call xj an outlier.
If Tn calculated is smaller than Tn as given in Table B.1, we say xj is not an outlier.
Table B.1 — Rejection values for Grubbs’ method
Sample 5 % significance Sample 5 % significance
size level size level
N (one-sided) N (one-sided)
3 1,150 20 2,56
4 1,46 21 2,58
5 1,67 22 2,60
6 1,82 23 2,62
7 1,94 24 2,64
8 2,03 25 2,66
9 2,11 30 2,75
10 2,18 35 2,82
11 2,23 40 2,87
12 2,29 45 2,92
13 2,33 50 2,96
14 2,37 60 3,03
15 2,41 70 3,09
16 2,44 80 3,14
17 2,47 90 3,18
18 2,50 100 3,21
19 2,53
© BSI 07-1999 31
BS ISO TR 5168:1998
B.3 Example
In the following sample of 40 values which are deviations from an average,
26 79 58 24 1 – 103 – 121 – 220
– 11 – 137 120 124 129 – 38 25 – 60
– 148 – 52 – 216 – 12 – 56 89 8 – 29
– 107 20 9 – 40 40 2 10 166
126 – 72 179 41 127 – 35 334 – 555
the suspected outliers are 334 and – 555.
To illustrate the calculations for determining whether – 555 is an outlier from Figure B.2:
Mean, x = 1,125
Experimental standard deviation, s = 140,813 6
Sample size, N = 40
From Table B.1 using Grubbs’ method for N = 40 at a 5 % level of significance (one-sided),
Tn,table = 2,87
Therefore, since 3,95 > 2,87, i.e.
Tn,calc > Tn,table
– 555 is an outlier according to the Grubbs’s test.
Table B.2 gives the results of this and two further iterations. The two suspected outliers, – 555 and 334,
are rejected by the Grubbs’ test.
Table B.2 is a normal probability plot of these data with the suspected outliers indicated. In this case, the
engineering analysis indicated that the – 555 and 334 readings were outliers, agreeing with the Grubbs’
test results.
Table B.2 — Results of Grubbs’ test
Sample size Experimental Mean
Suspected outlier standard deviation Calculated Tn Table Tn for P= 5
N s x
– 555 40 140,8 1,125 3,95 2,87
334 39 109,6 15,385 2,91 (stop) 2,86
– 220 38 97,5 7,000 2,33 2,85
32 © BSI 07-1999
Figure B.2 — Result of outlier tests
Annex C (informative)
Examples of estimation of uncertainty in airflow measurement
This annex contains two examples of fluid flow measurement uncertainty analysis. The first deals with
airflow measurement for an entire facility (with several test stands) over a long period. It also applies to a
single test with a single set of instruments. The second example demonstrates how comparative
development tests can reduce the uncertainty of the first example.
C.1 General
Airflow measurements in gas turbine engine systems are generally made with one of three types of
flowmeters: Venturis, nozzles and orifices. Selection of the specific type of flowmeter to use for a given
application is contingent upon a trade-off between measurement accuracy requirements, allowable
pressure drop and fabrication complexity and cost.
Flowmeters may be further classified into two categories: subsonic flow and critical flow. With a critical
flowmeter, in which sonic velocity is maintained at the flowmeter throat, mass flowrate is a function only
of the upstream gas properties. With a subsonic flowmeter, where the throat Mach number is less than
sonic, mass flowrate is a function of both upstream and downstream gas properties.
Equations for the ideal mass flowrate through nozzles, Venturis and orifices are derived from the
continuity equations:
q = ÔAV . . . (C.1)
In using the continuity equation as a basis for ideal flow equation derivations, it is normal practice to use
the principles of conservation of mass and energy and to assume one-dimensional isentropic flow.
Expressions for ideal flow will not yield actual flow, since actual conditions always deviate from ideal. An
empirically determined correction factor, the discharge coefficient (C) is used to adjust ideal to actual flow:
C = qactual/qideal . . . (C.2)
C.2 Example 1: Test facility

C.2.1 Definition of the measurement process
What is the airflow measurement capability of a given test facility? This question might relate to a
guarantee in a product specification or a research contract. Note that this question implies that many test
stands, sets of instrumentation and calibrations over a long period of time should be considered.
© BSI 07-1999 33
BS ISO TR 5168:1998
The same general uncertainty model is applied in the second example to a single stand process, the
comparative test.
These examples will provide, step by step, the entire process of calculating the uncertainty of the airflow
parameter. The first step is to understand the defined measurement process and then identify the source
of every possible error. For each measurement, uncertainties associated with calibration will be discussed
first, then data acquisition, data reduction, and finally, propagation of these uncertainties to the calculated
parameter.
Figure C.1 depicts a sonic nozzle flowmeter installed in the inlet ducting upstream of a turbine engine
under test for this example.
When a Venturi flowmeter is operated at critical pressure ratios, the flowrate through the Venturi is a
function of the upstream conditions only and may be calculated from
. . . (C.3)
C.2.2 Measurement error sources

Each of the variables in equation (C.3) must be carefully considered to determine how and to what extent
uncertainties in the determination of the variable affect the calculated parameter. A relatively large
uncertainty in some will affect the final answer very little, whereas small uncertainties in others have a
large effect. Particular care should be taken to identify measurements that influence the fluid flow
parameters in more than one way.
In equation (C.3), upstream pressure and temperature (p1 and T1) are of primary concern. Error sources
for each of these measurements are: (1) calibration, (2) data acquisition and (3) data reduction.
C.2.2.1 Figure C.2 illustrates a typical calibration hierarchy. Associated with each comparison in the
calibration hierarchy is a possible pair of elemental uncertainties, a systematic uncertainty and an
experimental standard deviation. Table C.1 lists all the elemental uncertainties. Note that these elemental
uncertainties of inlet pressure, p, are not cumulative, e.g. B21 is not a function of B11. The systematic
uncertainties should be based on interlaboratory tests if available, otherwise, the judgement of the best
experts must be used. The experimental standard deviations are calculated from calibration history data
banks.
Figure C.1 — Schematic of sonic nozzle flowmeter installation

upstream of a turbine engine
34 © BSI 07-1999
Figure C.2 — Typical calibration hierarchy

Table C.1 — Calibration hierarchy error sources
Calibration Systematic Experimental standard Degrees
uncertainty deviation of
Pa Pa freedom
SL — ILS B11 = 69 s11 = 13,8 v11 = 10

ILS — TS B21 = 69 s21 = 13,8 v21 = 15
TS — WS B31 = 69 s31 = 13,8 v31 = 20
WS — MI B41 = 124 s41 = 36,5 v41 = 30
The experimental standard deviation for the calibration process is the root-sum-square of the elemental
experimental standard deviations, i.e.
. . . (C.4)
= 43,6 Pa
The number of degrees of freedom associated with s are calculated from the Welch-Satterthwaite formula
as follows:
. . . (C.5)
= 54
© BSI 07-1999 35
BS ISO TR 5168:1998
The systematic uncertainty for the calibration process is the root-sum-square of the elemental systematic
uncertainties, i.e.
. . . (C.6)
= 172 Pa
Data acquisition error sources for pressure measurement are listed in Table C.2.
Table C.2 — Pressure transducer data acquisition error sources
Systematic Experimental standard Degrees
Error source uncertainty deviation of
Pa Pa freedom
Excitation voltage B12 = 69 s12 = 34,5 v12 = 40

Electrical simulation B22 = 69 s22 = 34,5 v22 = 90
Signal conditioning B32 = 69 s32 = 34,5 v32 = 200
Recording device B42 = 69 s42 = 34,5 v42 = 10
Pressure transducer B52 = 69 s52 = 48,3 v52 = 100
Environmental effects B62 = 69 s62 = 69 v62 = 10
Probe errors B72 = 117 s72 = 48,3 v72 = 60
The experimental standard deviation for the data acquisition process is
. . . (C.7)
= 119 Pa
. . . (C.8)
= 77
The systematic uncertainty for the data acquisition process is
. . . (C.9)
= 206 Pa
A computer operates on raw pressure measurement data to perform the conversion to engineering units.
Errors in this process are called data reduction errors and stem from curve fits and computer resolution.
36 © BSI 07-1999
BS ISO TR 5168:1998
Computer resolution is the source of a small elemental uncertainty. Some of the smallest computers used
in experimental test applications have six digits’ resolution. The resolution error is ± 1 in 106. Even though
this error is probably negligible, consideration should be given to rounding-off and truncating errors.
Rounding-off results in a random uncertainty. Truncating always results in a systematic uncertainty
(assumed in this example).
Table C.3 lists data reduction error sources.
Table C.3 — Pressure measurement data reduction error sources
Systematic Experimental standard Degrees of

Error source uncertainty deviation freedom
Pa Pa
Curve fit B13 = 69 s13 = 0 v13
Computer resolution B23 = 6,89 s23 = 0 v23
The experimental standard deviation for the data reduction process is
. . . (C.10)
= 0,0
The systematic uncertainty for the data reduction process is
. . . (C.11)
= 69,3 Pa
The experimental sample standard deviation for pressure measurement then is
. . . (C.12)
or
. . . (C.13)
= 127 Pa
Degrees of freedom associated with the experimental standard deviation are determined as follows:
. . . (C.14)
© BSI 07-1999 37
BS ISO TR 5168:1998
or
. . . (C.15)
= 97
therefore t95 = 2
The systematic uncertainty for the pressure measurement is
. . . (C.16)
or
. . . (C.17)
= 277 Pa
Uncertainty for the pressure measurement is
. . . (C.18)
. . . (C.19)
C.2.2.2 The calibration hierarchy for temperature measurements is similar to that for pressure
measurements. Figure C.3 depicts a typical temperature measurement hierarchy. As in the pressure
calibration hierarchy, each comparison in the temperature calibration hierarchy may produce elemental
systematic and random uncertainties. Table C.1 lists temperature calibration hierarchy elemental
uncertainties.
Table C.4 — Temperature calibration hierarchy elemental errors
Calibration Systematic Experimental standard Degrees
uncertainty deviation of
K K freedom
SL — ILS B11 = 0,056 s11 = 0,002 v11 = 2

ILS — TS B21 = 0,278 s21 = 0,028 v21 = 10
TS — WS B31 = 0,333 s31 = 0,028 v31 = 15
WS — MI B41 = 0,378 s41 = 0,039 v41 = 30
38 © BSI 07-1999
Figure C.3 — Temperature measurement calibration hierarchy

The calibration hierarchy experimental standard deviation is calculated as
. . . (C.20)
= 0,056 K
Degrees of freedom associated with s1 are
. . . (C.21)
= 53 > 30
therefore t95 = 2
The calibration hierarchy systematic uncertainty is
. . . (C.22)
= 0,578 K
A reference temperature-monitoring system will provide an excellent source of data for evaluating both
data acquisition and reduction temperature random uncertainties.
© BSI 07-1999 39
BS ISO TR 5168:1998
Figure C.4 depicts a typical set-up for measuring temperature with chromel-alumel thermocouples.
Figure C.4 — Typical thermocouple channel
If several calibrated thermocouples are utilized to monitor the temperature of an ice-point bath,
statistically useful data can be recorded each time measurement data are recorded. Assuming that those
thermocouple data are recorded and reduced to engineering units by processes identical to those employed
for test temperature measurements, a stockpile of data will be gathered, from which data acquisition and
reduction uncertainties may be estimated.
For the purposes of illustration, suppose N calibrated chromel-alumel thermocouples are employed to
monitor the ice-bath temperature of a temperature-measuring system similar to that depicted by
Figure C.4. If each time measurement data are recorded, multiple-scan recordings are made for each of the
thermocouples, and if a multiple-scan average ( x ij ) is calculated for each thermocouple, then the
average ( x j ) for all recordings of the jth thermocouple is
. . . (C.23)
where Kj is the number of multiple-scan recordings for the jth thermocouple.

The grand average ( x ) is computed for all monitor thermocouples as
. . . (C.24)
The experimental standard deviation ( s x ) for the data acquisition and reduction processes is then
. . . (C.25)
= 0,009 4 K (assumed for this example)
40 © BSI 07-1999
BS ISO TR 5168:1998
The degrees of freedom associated with s x are
. . . (C.26)
= 200 (assumed for this example)
Data acquisition and reduction systematic uncertainties may be evaluated from the same ice-bath
temperature data if the temperature of the ice bath is continuously measured with a working standard
such as a calibrated mercury-in-glass thermometer. There the systematic uncertainty is the largest
observed difference between x and the temperature indicated by the working standard acquisition and
reduction process. In this example, it is assumed to be 0,56 K, i.e.:
B x = 0,56 K . . . (C.27)
Error sources accounted for by this method are:

a) ice-point bath reference (random);
b) reference block temperature (random);
c) recording system resolution;
d) recording system electrical noise;
e) analog-to-digital conversion;
f) chromel-alumel thermocouple voltage versus temperature curve-fit;
g) computer resolution.
Several uncertainties which are not included in the monitoring system statistics are:
a) conduction uncertainty (BC);
b) radiation uncertainty (BR);
c) recovery uncertainty (BY);
d) calibration uncertainty (B1).
These uncertainties are a function of probe design and environmental conditions. Detailed treatment of
these uncertainties is beyond the scope of this work.
The experimental standard deviation for temperature measurements in this example is
. . . (C.28)
= 0,11 K
where
s1 = calibration hierarchy experimental standard deviation;
s x = data acquisition and reduction experimental standard deviation.
© BSI 07-1999 41
BS ISO TR 5168:1998
The degrees of freedom associated with sT are
. . . (C.29)
= 249, therefore t95 = 2
When v is less than 30, t95 is determined from a Student’s t table at the value of v. Since vT is greater
than 30 here, use t95 = 2.
Systematic uncertainties for the measurements are
. . . (C.30)
= 0,805 K
where
B1 = calibration hierarchy systematic uncertainties;
B x = data acquisition and reduction systematic uncertainties;
BC = conduction error systematic uncertainties (negligible in this example);
BR = radiation systematic uncertainties (negligible in this example);
BY = recovery factor systematic uncertainties (negligible in this example).
Uncertainty for the temperature measurement is
. . . (C.31)
= 1,03 K = 0,83 K
C.2.2.3 To minimize the uncertainty in the discharge coefficient, it should be calibrated using primary
standards in a recognized laboratory. Such a calibration will determine a value of C and the associated
systematic uncertainty and experimental standard deviation.
When an independent flowmeter is used to determine flowrates during a calibration for C, dimensional
uncertainties are effectively calibrated out. However, when C is calculated or taken from a standard
reference, uncertainties due to the measurement of pipe and throat diameters will be reflected as
systematic uncertainties in the flow measurement.
Dimensional uncertainties in large Venturis, nozzles and orifices may be negligible. For example, an error
of 0,001 mm in the throat diameter of a 5 mm critical flow nozzle will result in a 0,04 % systematic
uncertainty in airflow. However, these uncertainties can be significant at smaller diameters.
42 © BSI 07-1999
BS ISO TR 5168:1998
C.2.2.4 Nonideal gas behaviour and changes in gas composition are accounted for by selection of the proper
values for compressibility factor (Z), molecular weight (M) and ratio of specific heats (¾) for the specific gas
flow being measured.
When values of ¾, and Z are evaluated at the proper pressure and temperature conditions, airflow
uncertainties due to uncertainties of ¾, and Z will be negligible.
For the specific case of airflow measurement, the main factor contributing to variation of composition is the
moisture content of the air. Though small, the effect of a change in air density due to water vapour on
airflow measurement should be evaluated in every measurement process.

C.2.2.5 The thermal expansion correction factor (Fa) corrects for changes in throat area caused by changes
in flowmeter temperature.
For steels, a 17 K flowmeter temperature difference, between the time of a test and the time of calibration,
will introduce an airflow uncertainty of 0,06 % if no correction is made. If flowmeter skin temperature is
determined to within 3 K and the correction factor applied, the resulting uncertainty in airflow will be
negligible.
C.2.3 Propagation of uncertainty to airflow
For an example of propagation of uncertainties in airflow measurement using a critical-flow Venturi,
consider a Venturi having a throat diameter of 0,554 m operating with dry air at an upstream total
pressure of 88 126 Pa and an upstream total temperature of 265,9 K.
Equation (C.32) [identical to (C.3)] is the flow equation to be analysed:
. . . (C.32)
where
Assume, for this example, that the theoretical discharge coefficient (C) has been determined to be 0,995.
(To illustrate the uncertainty methodology, we will assume an experimental standard deviation of 0,000 5
in addition to a systematic uncertainty of 0,003.) Further assume that the thermal expansion correction
factor (Fa) and the compressibility factor (Z) are equal to 1,0. Table C.5 lists measured values, systematic
uncertainties, experimental standard deviations and degrees of freedom for each error source in the above
equation.
Note that, in Table C.5, airflow uncertainties resulting from uncertainties in Fa, Z, ¾, M and R are
considered negligible.
Table C.5 — Airflow measurement uncertainty sources
Error Measured Systematic of Uncertainty Uncertainty
Units standard
source value uncertainty freedom UADD URSS
deviation
v
p1 Pa 88 126 277 127 96 531 376
T1 K 266 0,8 0,11 250 1,02 0,83
d m 0,554 2,54 × 10–5 2,54 × 10–5 100 7,62 × 10–5 5,68 × 10–5
C 0,995 0,003 0,000 5 — 0,003 0,003 2
Fa 1,0 — — — — —
Z 1,0 — — — — —
¾ 1,40 — — — — —
M kg/kmol 28,9 — — — — —
R J/(K·kmol) 8,314 × 103 — — — — —
© BSI 07-1999 43
BS ISO TR 5168:1998
From equation (C.32), airflow is calculated as
= 52,3 kg·s–1
Taylor’s series expansion of equation (C.32) with the assumptions indicated yields equations (C.33)
and (C.34), from which the flow measurement experimental standard deviation and systematic
uncertainties are calculated.
. . . (C.33)
. . . (C.34)
By using the Welch-Satterthwaite formula, the degrees of freedom for the combined experimental standard
deviation are determined from
. . . (C.35)
44 © BSI 07-1999
BS ISO TR 5168:1998
which results in an overall degree of freedom > 30, and, therefore, a value of t95 of 2,0.
Total airflow uncertainty is then:
. . . (C.36)
. . . (C.37)
C.3 Example 2: Comparative test

C.3.1 Definition of the measurement process
The objective of a comparative test is to determine with the smallest measurement uncertainty the net
effect of a design change, such as a new part. The first test is performed with the standard or baseline
configuration. A second test, identical to the first except that the design change is substituted in the
baseline configuration, is then carried out. The difference between the measurement results of the two tests
is an indication of the effect of the design change.
As long as we only consider the difference or net effect between the two tests, all the fixed, constant,
systematic errors will cancel out. The measurement uncertainty is composed of random uncertainties only.
For example, assume we are testing the effect on the gas flow of a centrifugal compressor from a change to
the inlet inducer. At constant inlet and discharge conditions, and constant rotational speed, will the gas
flow increase? If we test the compressor with the old and new inducers and take the difference in measured
airflow as our defined measurement process, we obtain the smallest uncertainty. All the systematic errors
cancel. Note that, although the comparative test provides an accurate net effect, the absolute value (gas
flow with the new inducer) is not determined or, if calculated, as in Example 1, it will be inflated by the
systematic uncertainties. Also, the small uncertainty of the comparative test can be significantly reduced
by repeating it several times.
C.3.2 Measurement error sources
All errors result from random errors in data acquisition and data reduction. Systematic errors and hence
systematic uncertainties are effectively zero. Random uncertainty values are identical to those in
Example 1, except that calibration random uncertainties are classed as systematic uncertainties and,
hence, become effectively zero.
C.3.2.1 Comparative tests shall use the same test facility and instrumentation for each test. All calibration
errors are systematic and cancel out in taking the difference between the test results.
B1 = 0
and
s1 = 0, sC = 0
© BSI 07-1999 45
BS ISO TR 5168:1998
C.3.2.2
sp = s2
= 119 Pa [see equation (C.7)]
vp = v2 = 77 [see equation (C.8)]
sT = s x
= 0,094 K [see equation (C.25)]

vT = v x
= 200 [see equation (C.26)]
C.3.2.3 The test result is the difference in flow between two tests.
[see equations (C.36) and (C.37)]
C.3.2.4 Note that the differences shown in Table C.6 are entirely due to differences in the measurement
process definitions. The same fluid flow measurement system might be used in both examples. The
comparative test has the smallest measurement uncertainty, but this uncertainty value does not apply to
the measurement of absolute level of fluid flow, only to the difference.
Table C.6 — Uncertainty comparisons of Examples 1 and 2
Example 1 facility Example 2 facility
1) Experimental standard deviation, kg·s–1 (s) 0,079 0,076

2) Degrees of freedom (v) > 30 > 30
3) Systematic uncertainty, kg·s–1 (B) 0,246 0
4) Uncertainty, kg·s–1 0,40 0,22
46 © BSI 07-1999
BS ISO TR 5168:1998
C.4 Airflow example

In this example, airflow is determined by the use of a sonic nozzle and measurements of upstream
stagnation temperature and stagnation pressure (Figure C.5).
Figure C.5 — Flowrate through a sonic nozzle
The flowrate is calculated from
. . . (C.38)
where
q is the mass flowrate of air;
Fa is the factor to account for thermal expansion of the Venturi;
A is the Venturi throat area;
p1t is the upstream stagnation pressure;
T1t is the upstream stagnation temperature;
Î* is the factor to account for the properties of the air (critical flow constant);
C is the discharge coefficient.
The experimental standard deviation for the flowrate (sq) is calculated using the Taylor’s series expansion.
Assuming C equals 1 and has negligible uncertainty
. . . (C.39)
where
denotes the partial derivative of q with respect to Fa.
. . . (C.40)
By inserting the measured values and standard deviations from Table C.7 into equation (C.40), the
standard deviation of 0,16 kg·s–1 for airflow is obtained.
© BSI 07-1999 47
BS ISO TR 5168:1998
The systematic uncertainty in the flowrate calculation is propagated from the systematic uncertainties of
the measured variables. Using the Taylor’s series formula gives
. . . (C.41)
For this example,
. . . (C.42)
Taking the necessary partial derivatives gives
. . . (C.43)
By inserting the measured values and systematic uncertainties of the measured parameters from
Table C.7 into equation (C.43), a systematic uncertainty of 0,46 kg·s–1 is obtained for a nominal airflow
of 113 kg·s–1.
Table C.7 contains a summary of the measurement uncertainty analysis for this flow measurement. It
should be noted the uncertainties listed only apply to the nominal values.
Table C.7 — Flowrate data
Experimental standard Systematic
Parameter Unit Measured value
deviation uncertainty
Fa — 1,00 0,0 0,001

C — 1,0 0,0 0,0
Î* kg·K1/2·N –1·s–1 0,040 4 0,0 4,04 × 10–5
A m2 0,191 0,0 6,85 × 10–4
p1t Pa 2,54 × 105 345,0 345,0
T1t K 303,0 0,17 0,17
%q kg·s–1 113 0,16 0,46
Annex D (informative)
Examples of estimating uncertainty in open channel flow measurement
D.1 General
Evaluation of the overall uncertainty of a flow in an open channel will be demonstrated by considering (1)
the velocity-area method and (2) the weir method.
The method of measuring the flow is such that it is impractical to eliminate interdependent variables from
the equation before estimating flow uncertainty. Therefore, it involves evaluation of the interdependent
uncertainties specified in 7.4. In addition, measurement conditions often make it impossible to obtain the
replicate measurements needed for evaluation of experimental standard deviations. Under these
conditions, it is appropriate to assume that all the random uncertainties are equivalent to two
experimental standard deviations. Under this assumption, the random uncertainties can be propagated
with each other by means of the same root-sum-square formulas as the systematic uncertainties
[see equations (20) and (23)].
48 © BSI 07-1999
BS ISO TR 5168:1998
D.2 Example 1: Velocity-area method

D.2.1 Equation for discharge in an open channel (velocity-area)
The channel cross-section under consideration is divided into segments by m verticals. The breadth, depth
and mean velocity associated with any vertical i are denoted by bi, di and v i respectively (see Figure D.1).
The product q i = b i d i v i represents an approximation to the discharge (volumetric flowrate) in the ith
segment. The sum over all segments,
. . . (D.1)
represents an estimated or observed value of the total discharge.

If x and y are respectively horizontal and vertical coordinates of all the points in the cross section, and A is
its total area, then the precise mathematical expression for Q, the true volumetric flowrate (discharge)
across the cross-sectional area, can be written as
. . . (D.2)
The true discharge and the observed discharge are related by a proportionality factor representing the
approximation of the integral equation (D.2) by the finite sum equation (D.1), thus:
. . . (D.3)
where
In practice, Fm can be evaluated from analysis of measurements in which m is sufficiently large for the
effects on Qo of omitting verticals, in stages, to be determined. Fm is subject to a random uncertainty.
It may be convenient in practice to take an Fm variation with m that is a mean value of values for sections
of several different rivers, taken together. Then the actual variations of Fm from river to river, as compared
with the meaned variation, will involve both systematic and random uncertainties.
Fm is dependent on the number of verticals m, and tends to unity as m increases without limit. Thus,
equation (D.3) can be written approximately as
. . . (D.4)
with increasing accuracy as m increases.

This last form is the one that is given in ISO 748.
© BSI 07-1999 49
BS ISO TR 5168:1998
Figure D.1 — Schematic of velocity-area method of discharge measurement
(midsection method)
D.2.2 Overall uncertainty of the flow determination
It is plausible to assume that, at a given m, Fm and Qo can be treated as independent variables.
However, the qi in principle are not independent of one another, since the value corresponding to any one
vertical will be related to the values of adjacent verticals. Furthermore, there is an interdependence
between the di and v i corresponding to any particular vertical. Thus, by applying the principles for
combining experimental standard deviations (see clause 7), the following expression for sQ, the
experimental standard deviation of Q, can be derived from equation (D.3).
. . . (D.5)
where sij arise from the interdependence between qi and qj and s dvi from the interdependence between di
and v i .
It is convenient to introduce the notation s½ for relative experimental standard deviation. Thus s bi /bi is
written s′b , s F /F m is written s′F and, neglecting sij and s dvi , equation (D.5) becomes
i m m
. . . (D.6)
If the relative experimental standard deviations s′bi are all nearly enough equal, of value s9 b, and similarly
for the s′di and s′vi , then
. . . (D.7)
If the verticals are so located that qi . Qo/m, then
. . . (D.8)
50 © BSI 07-1999
BS ISO TR 5168:1998
In multipoint velocity-area methods, velocity is measured at several points on a vertical, and the mean
value is obtained by graphical integration or as a weighted average. The latter treatment can be expressed
mathematically as
. . . (D.9)
where the wp are constant weighting factors. The subscript i that identifies the particular vertical is
omitted to simplify the symbolism. The weighting factors usually are chosen so that . This
equation can also represent the single-point method, by taking k = 1.
In all cases, the estimates v so computed are subject to errors. These errors are due to improper placement
of the meter at depth, and to deviations of the actual velocity profile from the presumed profile. The effect
of these errors can be expressed by means of a multiplicative coefficient P analogous to the coefficient Fm
used for similar purposes in equation (D.3). The same analysis that led to equation (D.5) then yields the
following expression for relative experimental standard deviation of the average velocity v :
. . . (D.10)
in which s½ denotes relative experimental standard deviation in the subscript variable, V is measured point
velocity, and the ratio of wv-sum expresses the variability of weighted velocity over the depth of the
vertical. For a uniform k-point velocity profile, this ratio would equal 1/k. For an extremely nonuniform
profile, in which a single term dominated all the others, the ratio would equal 1. The latter value is adopted,
at least from small k values, for the sake of conservatism, with the result
. . . (D.11)
This choice also helps to represent the effect of any unaccounted-for correlations among point-velocity
errors in the same vertical.
In practice, the random uncertainty in the velocity measurement at a point is assumed to be due to a
meter-calibration relative experimental standard deviation, s9 c, together with a stream-pulsation relative
experimental standard deviation, s9 e. Then the relative experimental standard deviation for point
velocities is
The corresponding relative experimental standard deviation for average velocity in the vertical is
. . . (D.12)
D.2.3 Example of calculation of uncertainty

It is required to calculate the uncertainty in a current-meter gauging from the following particulars:
Number of verticals used 20
Exposure time of current meter at each point
in the vertical 3 min
Number of points taken on the vertical (single point,
two points, etc.) 2
Type of current-meter rating (individual or group) Individual
Average velocity in measuring section above 0,3 m·s–1
Details of the procedure are described in ISO 748.
© BSI 07-1999 51
BS ISO TR 5168:1998
The random and systematic uncertainties are combined by the root-sum-square and linear additive
methods as stated in 8.3, i.e. if 2 S9 Q and B9 Q are the percentage overall random and systematic
relative uncertainties respectively, then U9 Q, the percentage uncertainty in the current-meter gauging, is
D.2.3.1 The equation used for evaluating the overall relative experimental standard deviation is
[see equation (D.8)]
where
s9Q is the overall percentage experimental standard deviation;
s′F is the percentage experimental standard deviation due to the limited number of verticals used;
m
s9b is the percentage experimental standard deviation in measuring the width of segments;
s9d is the percentage experimental standard deviation in measuring the depth of segments;
s′v is the percentage experimental standard deviation in estimating the average velocity in each
vertical
and [see equation (D.12)]
where
s9p is the percentage experimental standard deviation due to limited number of points taken in the
vertical
(in the present example the two-point method was used, i.e. at 0,2 m and 0,8 m from the surface
respectively);
s9c is the percentage experimental standard deviation of the current-meter rating (in the present
example an individual rating was used at velocities of the order of 0,30 m/s);
s9e is the percentage experimental standard deviation due to pulsations (uncertainty due to the
random fluctuation of velocity with time; the time of exposure in the present example was three
one-minute readings of velocity).
The percentage values of the above partial uncertainties at the 95 % confidence level are tabulated
in D.2.3.2.
The equation for calculating the overall systematic uncertainty is
where
B9 Q is the overall percentage systematic uncertainty in discharge;
B9 b is the percentage systematic uncertainty in the instrument measuring width;
B9 d is the percentage systematic uncertainty in the instrument measuring depth; and
B9 c is the percentage systematic uncertainty in the current-meter rating tank.
The systematic uncertainties in the current-meter gauging are confined to the instruments measuring
width, depth and velocity, and should be restricted to 1 % as shown in D.2.3.2.
52 © BSI 07-1999
BS ISO TR 5168:1998
D.2.3.2 The values of the elemental uncertainties affecting uncertainty in discharge are listed in Table D.1
as percentage uncertainties at the 95 % confidence level. The numerical values are taken from ISO 748. It
is recommended, however, that each user determine independently the values of the uncertainties for any
particular measurement.
Table D.1 — Elemental uncertainties affecting uncertainty in discharge
Relative random
Relative systematic
uncertainty
Error source Units (2s: 95 %) uncertainty
B9
2s9
Fm, number of verticals — 5,0 % —

b, segment width m 0,5 % 1,0 %
d, segment depth m 0,5 % 1,0 %
vp, number of points in the vertical m·s–1 7,0 % —
vc, current-meter calibration m·s–1 2,0 % 1,0 %
ve, current-meter exposure time m·s–1 5,0 % —
Then, the overall random uncertainty in discharge is given by
= 5,4%
The overall systematic uncertainty is
= 1,7%.
The combination of both random and systematic uncertainties then gives the overall percentage
uncertainty in discharge, U9Q.
U9QADD = BQ9 + 2s9 Q
= 1,7 + 5,4
= 7,1 %
= 5,7 %
D.2.3.3 The discharge measurement may be expressed in the following form:
Discharge, Q m3·s–1
(Combined) uncertainty, U9QRSS 5,7 %
(Combined) uncertainty, U9QADD 7,1 %
Random uncertainty, 2s9Q 5,4 %
9
Systematic uncertainty, BQ 1,7 %
Uncertainties calculated in accordance with this Technical Report.
© BSI 07-1999 53
BS ISO TR 5168:1998
D.3 Example 2: Weir measurement

D.3.1 Weir data
It is required to calculate the discharge and the uncertainty in discharge for a triangular profile weir, given
the following details (see Figure D.2):
Gauged head, h 0,67 m
Breadth of weir, b 10 m
Height of weir, p 1m
Coefficient of discharge, Cd 1,163
Coefficient of velocity, Cv 1,054
The discharge equation is

. . . (D.13)
Details of the procedure are described in ISO 4360.

D.3.2 Uncertainty equations
Taylor-series analysis of the discharge equation yields the following equations for relative experimental
standard deviation and systematic uncertainty:
and
. . . (D.14)
in which s9 and B9 denote percentages of the subscript variables.
Figure D.2 — Triangular profile weir
54 © BSI 07-1999
BS ISO TR 5168:1998
D.3.3 Evaluation of discharge and uncertainties

The values of the elemental uncertainties affecting this problem are listed in Table D.2 as uncertainties at
the 95 % confidence level. The numerical values are based on information given in ISO 4360. It is
recommended, however, that each user determine independently the values of the uncertainties for any
particular measurement.
Table D.2 — Error element values
Random error limit Systematic
Variable Units Nominal value (2s: 95 %) error limit

2s, 2s9 B, B9
h m 0,67 0,003 m 0,003 m
0,45 % 0,45 %
b m 10,00 0 0,01 m
0,1 %
Cd C v — 1,226 0,5 % 1,5 %
g m·s–2 9,81 0 0
Substitution of the nominal values into the discharge equation yields
= 11,46 m3 ·s–1
Evaluation of the random uncertainties yields
= 0,84 %
Evaluation of the systematic uncertainties yields
= 1,65 %
Combining the random and systematic uncertainties by the root-sum-square (RSS) and linear additive
methods yields
U9QADD = B9Q+ 2s9Q
= 1,65 + 0,84
= 2,49 %
= 1,85 %
D.3.4 Presentation of results
The discharge Q may be reported as follows:
Discharge, Q 11,46 m3·s–1

(Combined) uncertainty, U9QRSS 1,8 %
(Combined) uncertainty, U9QADD 2,5 %
Random uncertainty, 2s9Q 0,8 %
Systematic uncertainty, B9Q 1,6 %
© BSI 07-1999 55
BS ISO TR 5168:1998
Uncertainties calculated in accordance with this Technical Report.
Annex E (informative)
Example of flowrate measurement in circular pipes
The following notation is used in this annex.
Symbol Meaning
C Discharge coefficient
D Diameter (bore) of pipe
d Diameter of throat of flowmeter
D(t9o) Value of D at temperature t9o
d(to) Value of d at temperature to
Uy Uncertainty in the variable y
ps Static pressure
qm Mass flowrate
ReD Reynolds number based on the pipe bore
tr Test temperature
µD Coefficient of expansion for pipe
µd Coefficient of expansion for flowmeter
¶ Diameter ratio of flowmeter, = d/D
%p Differential pressure across flowmeter
¼ Expansibility coefficient
È Dynamic viscosity
Ä Isentropic exponent
Ô Density of fluid
The formula for flowrate through differential pressure devices will be used as an example of how the
uncertainty in a mass flowrate measurement may be estimated. This general formula is
. . . (E.1)
While all known sources of uncertainty are considered, some are sufficiently small to be omitted from the
final calculation.
E.1 General calculation
E.1.1 Identification and listing of independent sources of error
The quantities in the expression for qm are not independent of one another, and so each must be examined
individually to determine the independent variables from which it is derived. This examination may
involve more than one step.
For example, the systematic uncertainty in C arises from the systematic uncertainties in ¶ and ReD and
the systematic uncertainty BC in the experiments used to determine the dependence of C on these
variables. However both ¶ and ReD depend on D, so these variables must be further subdivided.
Table E.1 can be drawn up showing the sources of the various quantities in equation (E. 1).
56 © BSI 07-1999
BS ISO TR 5168:1998
BÈ, BC, BÄ, B¼ and BÔ are the systematic uncertainties associated with experiments which determined the
dependence of È, C, Ä, ¼ and Ô on their associated independent variables, and d(to) and D(t9o) are the values
of d and D at the temperatures to and t9o at which they were measured. The values of d and D at the test
temperature, tr, are given by the equations
It is important to include each variable every time it occurs, since the uncertainty of that variable will have
a different effect on the flowrate measurement on each occasion. Thus, for example, the temperature tr at
which the flowrate is measured occurs twice in Table E.1 in the row opposite ReD, since it affects both D
and È
Table E.1 — List of independent sources of error
1st 2nd 3rd
C ¶ to, t9o , tr, d(to), D(t9o), µd, µD,

ReD t9o , tr, D(t9o), µD, ps, tr, qm, BÈ
BC BC
¼ ¶ to, t9o , tr, d(to), D(t9o), µd, µD
%p %p
ps ps
Ä ps, tr, BÄ
B¼ B¼
(1 – ¶4)–1/2 ¶ to, t9o , tr, d(to), D(t9o), µd, µD,
d2 d to, tr, d(to), µd,
%p %p %p
Ô Ô ps, tr, BÔ
The various quantities given in the third step of Table E.1 will be assumed to be independent for the
purpose of this example. Two variables are regarded as independent if the uncertainty in one does not
contribute to the uncertainty in the other or if there is no quantity the uncertainty of which affects the
uncertainty in both the variables. Thus, for example, ¶ and ReD are not independent, since the uncertainty
in D contributing to the uncertainty in ReD also contributes to the uncertainty in ¶.
E.1.2 Determination of the nature of the errors
Random uncertainties involved in the determination of a particular independent quantity may be
“fossilized” into systematic uncertainties due to that quantity. Thus, for example, the uncertainty in to may
be random, but the use of a fixed value of to will introduce a constant systematic uncertainty to the flowrate
measurement, since the value of d at the temperature of the test will be in error by a fixed amount. It is
therefore essential to consider the effects of uncertainties in the dependent variables on the flowrate, and
not on the variables themselves.
The “fossilized” uncertainties may vary according to the experimental conditions, e.g. flowrate or time of
test, but in the present example only fixed fossilized uncertainties are considered.
At this stage it is possible to draw up a list of the independent variables — these are given in Table E.2,
together with the way in which they affect the flowrate.
© BSI 07-1999 57
BS ISO TR 5168:1998
Table E.2 — Nature of errors

Source Number Nature of uncertainty
of error of effects introduced to flowrate
to 4 systematic
t9o 4 systematic
tr 8 systematic and random
d(to) 4 systematic
D(t9o) 4 systematic
µd 4 systematic
µD 4 systematic
ps 4 systematic and random
%p 2 systematic and random
È 1 systematic
C 1 systematic
¼ 1 systematic
Ä 1 systematic
Ô 1 systematic
E.1.3 Estimation of systematic uncertainties
The magnitude of the various systematic uncertainties depends very much on the equipment used. It can
be seen from Table E.2 that there are many cases where the systematic uncertainty of an experimental
determination of a variable leads to a systematic uncertainty in the flowrate measurement. It is not
possible to say here which variables will introduce a constant systematic error in a known direction: each
case must be examined individually.
E.1.4 Computation of random uncertainties
This is done in a straightforward way by analysing the data in the tests leading to the flowrate
measurement.
E.1.5 Calculation of sensitivity coefficient
The sensitivity coefficient of d(to), for example, is given by
Reference to Table E.1 shows that this may be written
. . . (E.2)
Similar expressions may be obtained for each of the independent variables, and these expressions may be
simplified since many of the terms are partial derivatives of explicit functions. Thus, for example
. . . (E.3)
58 © BSI 07-1999
BS ISO TR 5168:1998
Evaluating the partial derivatives whenever possible in this way, and substituting these values into the
various formulae for sensitivity coefficients, one obtains the following relations for the present example:
. . . (E.4)
© BSI 07-1999 59
BS ISO TR 5168:1998
The following terms cannot be computed directly from the flowrate equation:
In these cases, either the numerical technique specified in 7.4 b) must be used or a functional relationship
between the appropriate variables found. In using the numerical technique, reference will usually have to
be made to tables and graphs relating the relevant variables in the appropriate standards.
E.2 Simplified numerical example
E.2.1 General
In many practical situations it is not necessary to break down the quantities in the expression for mass
flowrate, q, fully into independent quantities. In such situations considering only the variables which
appear directly in the flow equation will give a result which is not significantly different to the rigorous
procedure outlined in E.1. This approach is, however, permissible only if it is specifically recommended in
the International Standard dealing with the appropriate flowmeter of flow measurement technique, since
otherwise there is a danger of ignoring component sources of error which may make a significant
contribution to the final uncertainty.
The case considered will be the mass flowrate measurement of steam using an orifice plate with tappings
at D and D/2.
60 © BSI 07-1999
BS ISO TR 5168:1998
E.2.2 Test results

The following values will be used:
to = 293 K
t9o = 283 K
d(to) = 0,141 28 m
D(t9o) = 0,202 21 m
%p = 26,783 kPa
tr = 693 K
ps = 400 kPa
d = 0,142 24 m
D = 0,203 20 m
Ô = 1,232 1 kg/m3
¼ = 0,973 79
¶ = 0,700 0
ReD = 7,13 × 105
C = 0,608 47
q = 2,775 4 kg/s
E.2.3 Sensitivity coefficients
From equation (E.1) and using the equations derived in 7.4, the experimental standard deviation and the
systematic uncertainty of q, sq and Bq, are given by:
. . . (E.5)
Performing the partial differentiations for the various terms in equation (E.1) yields:
. . . (E.6)
© BSI 07-1999 61
BS ISO TR 5168:1998
E.2.4 Sources of uncertainty

From equation (E.1), random and systematic uncertainties of the following variables have to be considered:
C; ¼; ¶; d; Ô; %p; D; ps; tr.
(Although D does not appear directly in the equation, it is incorporated in ¶ and must therefore be included;
similarly it is assumed in this example that Ô is calculated from values of ps and tr instead of being
measured directly, and so they must also be included.)
E.2.5 Component experimental standard deviations
It may be assumed for this example that all component experimental standard deviations were calculated
with greater than 30 degrees of freedom so that the small sample method is unnecessary and 2 may be
substituted for t95 in the final calculation of random uncertainty.
E.2.5.1 Experimental standard deviation of %p
From a knowledge of the scatter of the points about the best fit during the calibration of the transducer,
and taking into account the pulsations present during the measurement of flowrate, the experimental
standard deviation of %p was taken as
s%p= 100 Pa
E.2.5.2 Experimental standard deviation of ps
Fluctuations present in the flow produced random oscillations in the manometer; the experimental
standard deviation due to this was estimated as
sPs = 200 Pa
E.2.5.3 Experimental standard deviation of tr
From a knowledge of the scatter of results obtained during the calibration of the thermocouple, the
experimental standard deviation of tr using this instrument was deduced to be
s t r = 2,0 K
E.2.5.4 Combined component experimental standard deviation of Ô
It is assumed in this example that Ô is determined from measured values of ps and tr. Therefore, since
density is proportional to ps/tr, the experimental standard deviation of Ô is given by
. . . (E.7)
From tables of physical properties it is found that
Therefore,
= 0,003 61 kg·m–3
E.2.6 Systematic uncertainties

E.2.6.1 Systematic uncertainty of %p
The calibration of the pressure transducer used to measure %p showed that the systematic uncertainty was
0,5 % of reading, so for this measurement
B%p = 134 Pa
E.2.6.2 Systematic uncertainty of ps
Calibration showed that the manometer had a systematic uncertainty of 0,7 % of reading, so for this
measurement
B p = 2 800 Pa
s
62 © BSI 07-1999
BS ISO TR 5168:1998
E.2.6.3 Systematic uncertainty of tr

The maximum systematic uncertainty B t was taken as 2 K.
r
E.2.6.4 Systematic uncertainty BC of C
The systematic uncertainty is given by the appropriate standard as 0,7 %.
BC = 0,004 3
E.2.6.5 Systematic uncertainty B¼ of ¼
The systematic uncertainty is given in the appropriate standard as 4%p/ps or 0,268 % in this case. Hence
B¼ = 0,002 61
E.2.7 Combined component systematic uncertainties
E.2.7.1 Systematic uncertainty of d
Eight orifice diameters were measured, the uncertainty of the resulting values being 0,06 mm. The
measuring equipment had a systematic uncertainty of 0,02 mm. For the purposes of this example, the
uncertainty of d may be assumed to be the same as d(to), giving a total of
Bd = (0,062 + 0,022)1/2
= 0,063 mm or 6,3 × 10–5 m
E.2.7.2 Systematic uncertainty of D

Eight diameters were measured, the uncertainty of the results being 0,50 mm. The measuring equipment
had a basic systematic uncertainty of 0,05 mm. For the purposes of this example, the uncertainty of D may
be assumed to be the same as D(t9o ). One then obtains
BD = (0,52 + 0,052)1/2
= 0,5 mm or 5 × 10–4m
E.2.7.3 Systematic uncertainty of ¶

Since ¶ = d/D, the systematic uncertainty of ¶ is given by
. . . (E.8)
A comparison of the relative uncertainties of ¶ and D shows that B¶ may be approximated by BD. This
approximation further simplifies the final calculation of systematic uncertainty.
E.2.7.4 Systematic uncertainty of Ô
The systematic uncertainty for p found from tables is in this case 0,3 %, or 0,003 7 kg·m–3. It is assumed in
this example that Ô is determined from measured values of ps and tr. Therefore, since density is
proportional to ps/tr, the systematic uncertainty of Ô is given by
. . . (E.9)
= 0,010 1 kg·m–3
© BSI 07-1999 63
BS ISO TR 5168:1998
E.2.8 Combination of uncertainties

E.2.8.1 Combination of experimental standard deviations
From equation (E.5), the experimental standard deviation in the flowrate measurement arises from the
random contributions due to %p, ps and tr and is given, in percentage terms, by
. . . (E.10)
= 0,237 %
E.2.8.2 Combination of systematic uncertainties

Equation (E.5) may be written in percentage terms as follows:
. . . (E.11)
= 0,91 %
E.2.9 Presentation of results

The flowrate measurements may be expressed in one of the following forms:
1) Mass flowrate q = 2,775 kg·s–1

Systematic uncertainty B9q = 0,91%
Random uncertainty 2s9q = 0,47 %
Uncertainty U9RSS = 1,0 %
Uncertainty U9ADD = 1,4 %
2) Mass flowrate q = 2,775 kg·s–1

Systematic uncertainty Bq = 0,025
Random uncertainty 2sq = 0,013
Uncertainty URSS = 0,028
Uncertainty UADD = 0,038
Include the phrase “Uncertainties calculated in accordance with ISO/TR 5168”.
64 © BSI 07-1999
BS ISO TR 5168:1998
Annex F (informative)
Bibliography
[1] ISO 748:1997, Measurement of liquid flow in open channels — Velocity-area methods.
[2] ISO 772:1996, Hydrometric determinations — Vocabulary and symbols.
[3] ISO 3534-1:1993, Statistics — Vocabulary and symbols — Part 1: Probability and general statistical
terms.
[4] ISO 3534-2:1993, Statistics — Vocabulary and symbols — Part 2: Statistical quality control.
[5] ISO 3534-3:1985, Statistics — Vocabulary and symbols — Part 3: Design of experiments.
[6] ISO 4006:1991, Measurement of fluid flow in closed conduits — Vocabulary and symbols.
[7] ISO 4360:1984, Liquid flow measurement in open channels by weirs and flumes — Triangular profile
weirs.
© BSI 07-1999
BS ISO TR
5168:1998
BSI — British Standards Institution
BSI is the independent national body responsible for preparing
British Standards. It presents the UK view on standards in Europe and at the
international level. It is incorporated by Royal Charter.
Revisions
British Standards are updated by amendment or revision. Users of

British Standards should make sure that they possess the latest amendments or
editions.
It is the constant aim of BSI to improve the quality of our products and services.
We would be grateful if anyone finding an inaccuracy or ambiguity while using
this British Standard would inform the Secretary of the technical committee
responsible, the identity of which can be found on the inside front cover.
Tel: 020 8996 9000. Fax: 020 8996 7400.
BSI offers members an individual updating service called PLUS which ensures
that subscribers automatically receive the latest editions of standards.
Buying standards
Orders for all BSI, international and foreign standards publications should be
addressed to Customer Services. Tel: 020 8996 9001. Fax: 020 8996 7001.
In response to orders for international standards, it is BSI policy to supply the

BSI implementation of those that have been published as British Standards,
unless otherwise requested.
Information on standards
BSI provides a wide range of information on national, European and

international standards through its Library and its Technical Help to Exporters
Service. Various BSI electronic information services are also available which give
details on all its products and services. Contact the Information Centre.
Tel: 020 8996 7111. Fax: 020 8996 7048.
Subscribing members of BSI are kept up to date with standards developments

and receive substantial discounts on the purchase price of standards. For details
of these and other benefits contact Membership Administration.
Tel: 020 8996 7002. Fax: 020 8996 7001.
Copyright
Copyright subsists in all BSI publications. BSI also holds the copyright, in the
UK, of the publications of the internationalstandardization bodies. Except as
permitted under the Copyright, Designs and Patents Act 1988 no extract may be
reproduced, stored in a retrieval system or transmitted in any form or by any
means – electronic, photocopying, recording or otherwise – without prior written
permission from BSI.
This does not preclude the free use, in the course of implementing the standard,
of necessary details such as symbols, and size, type or grade designations. If these
details are to be used for any other purpose than implementation then the prior
written permission of BSI must be obtained.
If permission is granted, the terms may include royalty payments or a licensing

agreement. Details and advice can be obtained from the Copyright Manager.
BSI Tel: 020 8996 7070.
389 Chiswick High Road
London
W4 4AL

Iso 5168 - 1998

Uploaded by

Copyright:

Available Formats

Iso 5168 - 1998

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Iso 5168 - 1998

Uploaded by

Copyright:

Available Formats

British Standard

A single copy of this

This is an uncontrolled copy.

ICS 17.120.01; 17.120.10

This British Standard reproduces verbatim ISO TR 5168:1998 and implements

This British Standard, having Amendments issued since publication

ISBN 0 580 27553 1

Table C.4 — Temperature calibration hierarchy elemental errors 38

© BSI 07-1999 iii

3 Definitions and symbols

Figure 1 — Measurement error

Figure 2 — Illustration of terms relating to errors and uncertainties

Figure 3 — Systematic error

overbar (–) Mean value (of a variable).

x Arithmetic mean of the data values; xi.

xi Value of x at the ith datum point.

Y Arithmetic mean of the n measurements of the variable Y.

4 General principles of measurement uncertainty analysis

4.2 Random error

Figure 4 — Uncertainty interval x ± U (see also Figure 2)

Figure 6 — Random error

4.6 Uncertainty of measurements

4.7 Propagation of measurement uncertainties to test result uncertainties

5 Identification and classification of elemental measurement error sources

b) errors of measurement or data acquisition error sources (see 5.5);

Figure 7 — Basic measurement calibration hierarchy

Table 1 — Calibration hierarchy error sources

SL — ILS B11 s11 v11

Table 2 — Data acquisition error sources

Excitation voltage B12 s12 v12

Figure 8 — Data acquisition system

5.7 Method error sources

6 Estimation and presentation of elemental uncertainties

b) If there are M redundant instruments or M redundant measurements and the parameter to be

NOTE Here ij is used differently from elsewhere in the text.

For elemental uncertainties, the interval from B+ to B– shall include zero.

6.4.3 Averaging to reduce random error

Figure 9 — Calibration should compensate for test meter systematic error

The experimental standard deviation of the calibration constant K is:

7 Combination and propagation of uncertainties

7.3 Combining elemental systematic uncertainties

The systematic uncertainty for the basic measurement process is

7.4 Propagation of measurement uncertainties

subsidiary quantity is most easily obtained in one of two ways.

Analogously, the relative (nondimensional) sensitivity coefficient, Û i9 , is

and the systematic uncertainty by:

Figure 10 — Measurement uncertainty interval (UADD); symmetrical systematic uncertainty

Figure 11 — Measurement uncertainty; nonsymmetrical systematic uncertainty and

Table 5 — Recommended corrective action if the predicted

• Improve calibration • Larger test sample

9.3 After-measurement analysis and data validity

Figure 12 — Three after-measurement uncertainty interval comparisons

9.4 Summary for reporting error

d) the uncertainty formula , the uncertainty interval within

and . No more than two

significant places should be reported. For small samples see Annex A.

Table 6 — Elemental error sources

If the test result is an average, x , based on a sample of size N,

As N is a known constant, the number of degrees of freedom of s x is the same as s, i.e.