FD X07 021 Métrologie Et Application de La Statistique

Machine Translated by Google
ISSN 0335-3931
FD X 07-021
October 1999
Classification index: X 07-021
ICS: 03.120.30; 17.020
Fundamental standards
Metrology and statistical applications
Help with the process for estimating and using the uncertainty
of measurements and test results
E: Fundamental standards — Metrology and application of the statistics —

Help to the process for the estimation and the use of the measurement and
test results uncertainty
D: Grundnormen — Metrologie und Statisticanwendungen — Hilfe bei der
Verfahrensweise für die Schätzung und den Einsatz der Meß- und
Prüfergebnisunsicherheit
Documentation
booklet
published by AFNOR in October 1999.
Correspondence At the date of publication of this document, there are no European works
reserved
AFNOR
rights
1999
or international documents dealing with the same subject.

All
—
©
Analysis This document provides assistance with the process for estimating the uncertainty of measurements
and test results.
It completes and expands the approved standard NF ENV 13005 “Guide for the expression
measurement uncertainty” and establishes a link between the approach presented in this
guide and the series of standards NF ISO 5725-1 to 6 “Application of statistics — Accuracy
(accuracy and reliability) of results and measurement methods”.
A significant part of the document is devoted to the presentation of examples
from various fields.
Descriptors International Technical Thesaurus: statistics, metrology, measurement, testing,

test result, estimation, error calculation, precision, accuracy, uncertainty.
Changes Corrections
Published and distributed by the French Standardization Association (AFNOR), Tour Europe 92049 Paris La Défense Cedex Tel. : 01 42 91 55 55
— Tel. international: + 33 1 42 91 55 55
© AFNOR 1999 AFNOR 1999 1st drawing 99-10-F

Metrology in the company AFNOR X07B
Members of the standardization commission

President: Mr BARBIER
Secretariat: M CLOAREC — AFNOR
M ALLIOUZ ESSO SAF

M ALVERNHE AERONAUTICS AND SPACE STANDARDIZATION BUREAU (BNAE)
M ANTOINE CENTRAL LABORATORY OF ELECTRICAL INDUSTRIES (LCIE)
M ARRIAT BUREAU VERITAS
MRS AUTICET SNCF
M BARBER AEROSPACE
MRS BAVELARD CERIB
MRS BERNAZZANI KODAK INDUSTRY
M BORREIL MINISTRY OF DEFENSE — DGA DCA CEAT
M BRIGODIOT AEROSPACE
M BRUNET AFNOR
MRS BUIL SOFIMAE SA
M BUSUTTIL GDF DION PRODUCT TRANSPORT CTO
M CHAILLIE PSA PEUGEOT CITROEN
M COLLAY NATIONAL HIGHER SCHOOL OF MINES
M CORDEBOIS DASSAULT AVIATION
M DABERT THOMSON CSF SERVICES INDUSTRY
MRS DE PALMA ECM METROLOGY
M DELAMASURE MECHANICAL STANDARDIZATION UNION (UNM)
M DEPINOY SCIENTIFIC AND TECHNICAL BUILDING CENTER (CSTB)
M DEVIGNES SNCF
M DUMONT SCHNEIDER ELECTRIC SA
M ERARD CENTRAL LABORATORY OF ELECTRICAL INDUSTRIES (LCIE)
M FOLLIOT BALLISTIC AND AERODYNAMIC RESEARCH LABORATORY (LRBA)
M FOURCADE CEA CESTA
MRS FRABOULET ANJOU RESEARCH
M GELY SOPEMEA SA
M HITIE GEMPLUS CARD
M KELLER NATIONAL BUREAU OF METROLOGY (BNM)
M KRYNICKI HEWLETT PACKARD FRANCE
M LARQUIER MCE CONSEIL
M LAULAGNET INERIS
M LE BECHEC CENTRAL LABORATORY OF BRIDGES AND ROADS (LCPC)
M LEGEAY CENTRAL LABORATORY OF BRIDGES AND ROADS (LCPC)
MRS LENAN E2M
M LEVEL SANOFI RESEARCH
M MAGANA MINISTRY OF INDUSTRY — DARPMI
M MARDELLE DASSAULT ELECTRONIQUE
M MARSCHAL NATIONAL TESTING LABORATORY (LNE)
M MARTINEZ CIM CONSULTANTS
M MICHAEL MINISTRY OF DEFENSE — DGA — DSTI — ETBS
M MILLERET SOMELEC SA
M MONAT CTIF
MRS MORIN AFNOR
M NAUDOT ALCATEL CIT
MRS NOTIS AFNOR
M ODRU UNPP
M PENIN CAST SA INSA
M PICHON AFNOR
M PINAUD THOMSON CSF SERVICES INDUSTRY
M PRIEL NATIONAL TESTING LABORATORY (LNE)
M PRIN MINISTRY OF DEFENSE — DGA CELAR
M RAMBAUD TEKTRONIX SA
M REGNAULT NATIONAL TESTING LABORATORY (LNE)
MRS FOX DOUAI MINES SCHOOL
M REST FRENCH ACCREDITATION COMMITTEE (COFRAC)
M ROBIN AFNOR
M SENELAER HIGHER SCHOOL OF METROLOGY
M SERVE TECHNICAL ELECTRICITY UNION (UTE)
M SOUCEK AOIP INSTRUMENTATION
M STAROPOLI GDF
M VALITECHK MEASUREMENT UNION
M VANHALWYN ROHDE AND SCHWARZ
M VILLAROYA APAVE PARISIENNE
M VULOVIC GDF DION RESEARCH CERMAP
Statistical methods AFNOR X06E
Members of the standardization commission

President: M PERRUCHET
Secretariat: MRS MONTOYA — AFNOR
M AUXERRE AFNOR
M BARBER AEROSPACE
MRS BONIFACE IUP IBLIS
MRS BAKER ROHM AND HAAS FRANCE SA
M BOULARAN GDF DION OF RESEARCH
MRS BOUVENOT AFNOR
M BRUNET AFNOR
M BRUNSCHWIG
M CAZALBOU FRANCE TELECOM DQF
M CHEROUTE FORESIGHT SYSTEMS
M KNIGHT SCHNEIDER ELECTRIC SA
M DAUDIN INAPG
M OF IRON AFNOR
MRS DESENFANT NATIONAL TESTING LABORATORY (LNE)
M FAUCHON PROMOSTAR SARL
M FEINBERG INAPG
M GALINDO RHODIA QUALITY SERVICES
M LAURENT FILTRAUTO SA
MRS LEMER ESSILOR INTERNATIONAL SA
MRS NEUILLY
MRS OUDIN DARRIBERE
M PALSKY RHODIA QUALITY SERVICES
M PARRUCHET UTAC
MRS PETETIN AMOVI
M REVY BERTRAND FAURE EQUIPEMENTS SA
M SADO TOTAL RAFFINAGE DISTRIBUTION SA
M SAPORTA CNAM
M SIXOU RENAULT
M WENISCH SQIFE
M ZANKEVICH DRIRE
Members of the “Measurement uncertainty” working group

Moderator: Mr PRIEL — National Testing Laboratory (LNE)
M BELLON CENTRAL LABORATORY OF ELECTRICAL INDUSTRIES (LCIE)

MRS BERNAZZANI KODAK INDUSTRY
M BOCK TOTAL EUROPEAN RESEARCH AND TECHNICAL CENTER
M BEVERAGE FRENCH ACCREDITATION COMMITTEE (COFRAC)
M BRIGODIOT AEROSPACE
M BRUNSCHWIG
MRS DESENFANT NATIONAL TESTING LABORATORY (LNE)
M MOSSE ALCATEL SPACE
M NIMANBEG RENAULT
M PARRUCHET UTAC
M POIRIER UTAC
M PRIEL NATIONAL TESTING LABORATORY (LNE)
M REST FRENCH ACCREDITATION COMMITTEE (COFRAC)
FD X 07-021 —4—
Summary
Page
Foreword .............................................. .................................................. .................................................. ... 5
1 Application domain .............................................. .................................................. ............................. 5
2 Normative references .................................................. .................................................. .................. 6
3 Terms and definitions .............................................. .................................................. ....................... 6
4 Approach for the estimation, expression and use of the uncertainty of measurements
and test results ......................... .................................................. ................. 8
5 Comments on the developed flowchart ......................................... ................................... 12
6 Appendices .................................................. .................................................. ......................................... 23
Annex A (informative) Estimation of the calibration uncertainty of a liquid expansion

thermometer ......................... .................................................. ................................... 24
Annex B (informative) Estimation of the uncertainty of a measurement of polluting CO emissions

in the automotive industry ............................................. .................................................. ......... 28
Annex C (informative) Estimation of uncertainty associated with height measurement

of the first bounce of a tennis ball ......................................... ......................................... 30
Annex D (informative) Estimation of the uncertainty of a current intensity measurement .................................. 34
Annex E (informative) Estimation of the uncertainty of a luminance coefficient measurement

materials for marking a roadway ......................................... ................................. 41
Annex F (informative) Estimation of uncertainty of sodium bromide determination

in a black-white developer ......................................... .................................................. ...... 47
Annex G (informative) Estimation of the uncertainty of a temperature measurement in an industrial environment 53
Annex H (informative) Estimation of the uncertainty of a measurement of the filterability limit temperature
fuels for diesel engines and domestic fuel oils ......................................... ...... 56
Bibliography .................................................. .................................................. .................................................. 58

—5— FD X 07-021
Foreword
Although
asthis
wellclassification
as that of statistics.
index document is related to the field of metrology, it
concerns was developedof
(He metrologists bythe
a group of statisticians from and
General Commission “Metrology in the Commission of
Standardization “Statistical methods” (X06E)).
In the absence of a possibility of double indices, the new classificationsystem of standards “International
Classification of Standards (ICS) adopted by the
for AFNOR
the catalog allows this document to be identified in
standards
collection of each of the two domains.
NF ENV 13005
standard “Guides
for ENV 13005 the expression
“Guides of measurement
the expression of the uncertainty” takes up the European prestandard
uncertainty of which itself takes themeasurement”
developed theby ISO Technical Advisory Group No. (TAG 4). text for
This provides the scientific metrological bases 4 text and for

assess the uncertainties of international consensus.
a known These bases supposed
received
to be are measurement,
of the reader therefore reproduced in this document.

and are not
In a testing laboratory
a one for testing aas in aofmetrology
result laboratory,
need to estimate assessing
the associated the quality
uncertainty. However,ofthere
a result
exists or measurement, we have a
difference between the results from a test of a nature or measurement.
A set of operations aimed atmeasurement

is having determiningvalue
a quantity (Definition
for one of 07-001).
the NF
tests. field of metrology In the in tions the we generally has information on the progress of the operas - more on
that case
A test the technical

equipment,
operationorganism,
to be determined
physical is
phenomenon,
which consists
process
of onespecified
or more operating
product, material,
mode (NF 45020). several features the performance
Or service given, depending on A

EN
is the equipment intended to represent The test laboratory conditions encountered site on or in service, to which the product can
duit or is submitted. dations, even often undergo deformation, damage, complete destruction (impact tests, etc.).
held at fire,
Eitherfirst
site that this on or in
approach, a generally independent method
in 45020), the of the (as defined metrology laboratory, result of a measurement is
measurement (length, mass, etc.), in a then test that case test result
standard NF EN depends
the one who on method, method present
A conventional character.
This documentation booklet presents a measurement of a test a detailed procedure without out
much uncertainty
an address to establish to justify estimate to carry
result. or uncertainties in test results. to estimates

He having and the people
measurementsor
NFcompletes
ENV 13005thestandard to be used when it is formalized. proposing a model based approach process of on interlaboratory tests which can
describing obtaining a test result measures it or can't born
it includesofthis
Finally, examples procedure application in different fields.
of the
The reader's attention difficulties
is the estimation of
of ameasurement
measurement ofuncertainties. Estimate
a test result requires uncertainty
good knowledge of
method, analysis, in order

of to
errors. collective work involving the operator control A causes
the mind or one to identify
the and the one which completely technical

It is very often a of testing.
the measure or
the uncertainties of
Finally, documentation the difficulties
booklet writers of wish
a mathematical
to raise order the estimation
awareness of theisdifficulties
never statistical,
of but only of test to evidence of the
Or
cause
difficulties
of errors.linked to are a correct analysis and setting process measurement or
1 Area of application
The proposed approach applies to all areas of measurement and testing when it is necessary to determine
an uncertainty associated with an outcome, particularly in meeting the corresponding requirements formulated
in the ISO 9000 and ISO 14000 family standards, in the EN 45001 standard, in the ISO/IEC 25 guide and on
the national plan in standard NF X 07-010.
FD X 07-021 —6—
2 Normative references
This document incorporates by dated or undated reference provisions from other publications. These normative references are cited at the appropriate
places in the text and the publications are listed below.
For dated references, subsequent amendments or revisions of any of these publications do not
apply to this document only if they have been incorporated by amendment or revision. For references no
dated, the latest edition of the publication referred to applies.
NF ENV 13005, Guide to Expression Uncertainty for Measurement (classification index: X 07-020) .
NF EN 30012-1, Quality assurance requirements for measuring equipment — Part 1: Confirmation

Metrological measuringequipment (classification index: X 07-009-1) .
NF EN 45001, General criteria for the operation of testing laboratories (classification index:
X 50-061) .
NF EN 45020, Standardization and related activities — General vocabulary (classification index: X 50-080) .
NF EN ISO 9001, Quality systems — Model quality assurance for production,installation in design, development, pro-
associated services and (classification index: X 50-131) .
NF EN ISO 9002, Quality systems — Quality assurance model for in production, installation and delivery
associated (classification index: X 50-132) .
NF EN ISO 9003, Quality systems — Quality assurance model for final inspections and tests (hint
classification: X 50-133) .
NF EN ISO 14001, Environmental management systems — Guidelines and specifications for its
use (classification index: X 30-200) .
NF EN ISO 14253-1, Geometric specification (GPS) — Verification of product parts and

by measurement
Rules decision compliance non-compliance with specific measuring instruments — Part 1: from the to the to prove or cification
(classification index: E 10-201-1) .
NF ISO 3534–1, and 1: and

Statistics — Symbol vocabulary — Part Probability general statistical terms
(classification index: X 06-002-1).
NF ISO 5725 (all parts), (classification index: Application of statistics — Accuracy (accuracy) of results and and
measurementmethods X 06-041-1 to 6) .
ISO/IEC Guide 25, General requirements concerning competence of calibration and testing laboratories.
NF ISO 10012-2, Insurance Measurement

quality of — Part director
master 2: equipment Lines there for the three measurement
process
(classification index: X 07-009-2).
NF ISO 14004, Environmentalmanagement systems — General guidelines regarding principles, systems, techniques and implementation
(classification index: X 30-204)
NF X 07-001, Core Standards — International Vocabulary general core terms of and

metrology (VIM).
NF X 07-010, Metrology — metrological function in the company.
3 Terms and definitions

For the purposes of this document, the following terms and definitions from the International Vocabulary of
fundamental and general terms of metrology (VIM) and NF ISO 3534-1, apply.
The reader's attention is drawn to the slight differences (without consequences) in the definitions of the terms
repeatability and reproducibility proposed in the International Vocabulary of Fundamental Terms and
general metrology (VIM, NF
— Vocabulary and symbols — Part 1: Probability and general statistical terms.
—7— FD X 07-021
3.1
repeatability (of measurement results) closeness of
agreement between the results of successive measurements of the same measurand, measurements carried out under all the same measurement
conditions
NOTE 1 These conditions are called repeatability conditions.
NOTE 2 Repeatability conditions include:
— same operating procedure;

— same observer;
— same measuring instrument used under the same conditions;
— same place;
— repetition for a short period of time.
NOTE 3 Repeatability can be expressed quantitatively using the dispersion characteristics of the results.
[VIM, § 3.6]
3.2
reproducibility (of measurement results) closeness of
agreement between the results of measurements of the same measurand, measurements carried out by varying the measurement conditions
NOTE 1 For an expression of reproducibility to be valid, it is necessary to specify the conditions that are varied.
NOTE 2 The conditions that are varied may include:
— measurement principle;
— measurement method;
- observer ;
- measuring tool ;
— reference standard;
- place ;
- Terms of use ;
- time.
NOTE 3 Reproducibility can be expressed quantitatively using the dispersion characteristics of the results.
NOTE 4 The results considered here are usually the corrected results.
[VIM, § 3.7]
3.3
precision
closeness of agreement between independent test results obtained under stipulated conditions
NOTE 1 The precision depends only on the distribution of random errors and has no relationship with the true value or the specified
value.
NOTE 2 The precision measure is expressed in terms of unfidelity and is calculated from the standard deviation of the test results.
Low fidelity is reflected by a large standard deviation.
NOTE 3 Independent test results mean results obtained in a manner not influenced by a previous result on the same or similar
equipment. Quantitative measures of fidelity depend critically on stipulated conditions. Repeatability and reproducibility conditions
are particular sets of stipulated extreme conditions.
[NF ISO 3534-1, § 3.14]
3.4
repeatability
fidelity under repeatability conditions [NF ISO 3534-1,
§ 3.15]
FD X 07-021 —8—
3.5
repeatability conditions
conditions where test results are obtained by the same method on identical test individuals in
the same laboratory, by the same operator, using the same equipment and during a short interval of time
NOTE See 3.14, NOTE 3.
[NF ISO 3534-1, § 3.16]
3.6
reproducibility
fidelity under conditions of reproducibility
[NF ISO 3534-1, § 3.20]
3.7
reproducibility conditions
conditions where test results are obtained by the same method on identical test individuals in
different laboratories, with different operators and using different equipment
[NF ISO 3534-1, § 3.21]
4 Approach to the estimation, expression and use of uncertainty

measurements and test results
The proposed approach includes five essential phases:
— Expression of the need for measurement or testing Phase A;

— Choice of resources Phase B
— Measuring or testing process Phase C
— Analysis and exploitation of results Phase D
- Decision Phase E
It is summarized in the simplified flowchart in Figure 1.
The different phases are described in the flowchart developed in Figure 2. Each of these stages
is broken down into modules referenced A1, A2, B1, etc. ; an explanatory text is provided for each module.
—9— FD X 07-021
NF ENV 13005
Figure 1 — Simplified flowchart

FD X 07-021 — 10 —
NF ISO 5725
NF ENV 13005
Figure 2 — Approach for estimating the uncertainty of measurements and test results
— 11 — FD X 07-021
Figure 2 — Approach for estimating the uncertainty of measurements and test results (end)
FD X 07-021 — 12 —
5 Comments on the developed flowchart

5.1 Phase A1: Definition of the need for measurement or testing
The measurement or test relates to a product: the result of the measurement or associated test constitutes decisive technical
data in the decision process that concerns it.
Objectives of the measurement or testing process:
— characterize a product with a view to its definition, its apprehension or its knowledge (sector of the
scientific research: example of dating measurements, etc.);
— compare the characteristics and performances of a product to its technical specifications (performance management in a
customer-supplier context); several types of measurements and tests appear during the product life cycle:
- measurements or development tests in design, product development; - measurements or

qualification tests; - measurements or
acceptance tests; - operational evaluation
measurements or tests; - follow-up measurements or
tests during use or operation;
— classify products based on a measurable physical characteristic; — control production
processes.
Definition of the measurement or test context:
— the product subject to measurement or testing;

— the measurand in terms of measurable physical quantity and its associated technical specification; — the
reference operating conditions relating to the environment of the product during its use and covering aspects such as climatic,
mechanical, electromagnetic, ecological conditions and the human factor (see 3.1, D.1 and D.2 of the NF ENV 13005
standard “Guide for the expression of the uncertainty of
measure".
Constraints:
— costs linked to:
- the availability of measurement and testing equipment (purchase, depreciation, rental, etc.); - connecting
equipment and maintaining their metrological characteristics; - control of operational conditions,
consumables and specific additional means; - the definition, then validation of specific procedures and methods; -
resource management; - support and archiving of data; - the time of installation and
operational implementation of
the measurement and testing means; - the time
taken to produce and use the results of measurements and tests; — the deadlines will depend on the planning
of the experimental tasks, the realization, exploitation and validation of the results of
measurements and tests;
— other constraints: they may be contractual, technical or regulatory, they are linked: - to the measurement or test
method imposed; - maximum tolerated errors; - the
presentation of a calibration certificate
from an accredited laboratory; - requirements for demonstrating quality assurance of
measurements and tests; - permitted or unauthorized alteration of the product; - special subcontracting conditions.
From these objectives and constraints we deduce an acceptable uncertainty.

— 13 — FD X 07-021
5.2 Phase B1: Choice of resources
The choice and definition of resources require consideration of the technical needs and economic and commercial conditions identified in
module A.1: — it is important to achieve the adequacy of the technical performance of the
equipment with the technological requirements of the company taking into account the constraints of implementation and use (connection,
maintenance, environment, etc.). The homogeneity of the instrument pool as well as the possibilities for upgrading the equipment are
criteria to be taken into consideration. Establishing specifications may be recommended for complex equipment;
— the economic and commercial conditions are to be determined jointly by the managers concerned by this investment (responsible for
measurements, responsible for the metrological function, fleet manager, etc.) in agreement with the purchasing manager, taking into
account the skills of the supplier to meet the requirements, depreciation conditions, maintenance contracts and technical assistance.
It may be considered to use subcontracting;
— the qualification of the technical staff involved, as well as the maintenance of their skills must be determined
mined in relation to human resources managers;
— the choice of measurement and testing means may be determined based on evaluations resulting from experience acquired in other
companies or carried out by metrology or testing laboratories or user associations. measuring and testing equipment. Technical
documentation for the equipment can help the company in its choice;
— the actors who intervene in this process are on the one hand the metrologist or the test manager and on the other
by the technicians and operators as well as the people mentioned above.
5.3 Phase B2: Preliminary analysis of the process

This phase allows experienced metrologists and test managers to save time. In particular, it is of great interest when one wants to acquire a
measuring instrument and it then makes it possible to make a forecast evaluation of the measurement uncertainty (see appendix G).
This analysis is not limited to examining the instrumentation but must also cover the environment, personnel, etc.
This phase is in principle not necessary, however, it turns out to be extremely important from a practical point of view. It can only be carried
out by a specialist in the field of measurement or testing considered and will call on his skills and professional experience.
In this phase, it is a question of making a rapid estimate (“order of magnitude”) of the uncertainty of the measurement result due to the
contributions of the different components of the uncertainty.
The result of this preliminary analysis allows:
— to know the different components and their relative weight, and thus to determine on which components
uncertainty requires stepping up efforts;
— to assess whether the resources envisaged can be suitable for the uncertainty objective set. In fact, this first approach makes it possible to
obtain an order of magnitude of the uncertainty. This value (roughly estimated) makes it possible to decide whether the envisaged process
is likely to be suitable and it should then be continued, or if it is doomed to failure, it is then appropriate to look for another measurement
or test process.
In certain cases the estimation of uncertainty can stop at this stage. Indeed, if the uncertainty thus evaluated turns out to be small compared
to the acceptable uncertainty, it is not necessary to continue the estimation and the calculation made during this preliminary analysis can serve
as justification.
5.4 Phase C1: Complete analysis of the process

The complete analysis of the measurement or testing process is undeniably the most difficult phase in uncertainty estimation. It will require
skill, physical sense, critical thinking on the part of the person responsible for estimating the uncertainty of the measurement or test result.
The objective of analyzing the process is to establish the way in which the result is obtained, as well as an assessment as complete as
possible of the causes of errors that may be introduced into the process.
FD X 07-021 — 14 —
Some of these errors are reduced:
— either by repeating the measurements (reduction of random errors); —
or by making corrections (reduction of systematic errors).
In order to establish as exhaustive a list as possible, the following list can be used (extracted from the standard
NF ENV 13005):
a) incomplete definition of the measurand;

b) imperfect realization of the definition of the measurand; c)
non-representative sampling, the measured sample may not represent the defined measurand; d) insufficient
knowledge of the effects of environmental conditions on measurement or imperfect measurement of environmental conditions;
e) observer bias when reading analog instruments; f) finite resolution of the

instrument or mobility threshold; g) inaccurate values of
standards and reference materials; h) inaccurate values of constants
and other parameters obtained from external sources and used in the data processing algorithm;
i) approximations and assumptions introduced into the measurement method and procedure; j) variations
between repeated observations of the measurand under apparently identical conditions.
We also examine the possibility of correlations between certain input quantities.
5.5 Phase C2: Process modeling

Modeling the process means transcribing in the form of mathematical writing the way in which all the information available to
the experimenter is used to calculate the measurement or test result that he announces. These are for example: a series of
readings from the instrument, the value of a correction read in a calibration certificate, the value of a quantity obtained in a
manual, the measurement or estimation of a magnitude of influence, etc.
The measurand Y is generally not measured directly but is determined from N other quantities X1, X2, …, XN through the
functional relation f, the model of the process is then:
Y = f(X1, X2, …, XN)
Among the Xi are the corrections (or corrective factors) made or not made, as well as quantities which take into account all
other sources of variability such as: different observers, instruments, samples, laboratories and measurement periods.
The function f therefore does not simply express a physical law, but the measurement or testing process and in particular, the function must
contain all the quantities which contribute significantly to the uncertainty of the final result.
When several input quantities Xi , Xj depend on the same quantity T they are correlated,, and it is sometimes useful to write
the mathematical model developed by explaining the input quantities concerned according to this same quantity T of way to
avoid the introduction of covariance terms.
EXAMPLE
Two masses of 50 g, A and B, are successively compared to the same standard E of 50 g.
A and B are then used simultaneously to make a 100 g standard.
What is the composite standard uncertainty uc(y) on this assembly of the two masses?
The two comparisons are modeled by the following equations:
A = E + X1
B = E + X2
The following Table 1 represents the two ways of modeling, depending on whether or not the covariance terms are highlighted.
— 15 — FD X 07-021
Table 1 — Process modeling
Model highlighting covariance terms Model developed in uncorrelated elements
Y=A+B Y = E + X1 + E + X2
application of the law of propagation application of the law of propagation
2 2
(y) = u2(a) + u2(b) + 2u(a, b) uc uc (y) = 4u2(e) + u2(x1) + u2(x2)
1)
u(a, b) = u2(e)
2
uc (y) = u2(e) + u2(x1) + u2(e) + u2(x2) + 2u2(e)
2 2
units (y) = 4u2(e) + u2(x1) + u2(x2) uc (y) = 4u2(e) + u2(x1) + u2(x2)
1) Application of formula (F.2) of appendix F of NF ENV 13005. standard
We see that following the development of the model, the application of the law of propagation of uncertainty makes
appear or not a covariance term.
5.6 Phase C3: Interlaboratory tests and comparisons

The test method can turn out to be very complex, which is why its implementation conditions are decisive and must be
perfectly mastered.
The quality of a test method is assessed by its accuracy: accuracy and precision (NF ISO 5725).
Accuracy is characterized by the difference between the average of a set of test results and the accepted reference
value.
Precision is defined as the ability of a method to provide test results that are very close to each other.
others, when the same product is tested several times under given conditions of application of the
method. Fidelity covers repeatability, reproducibility through intermediate fidelities. Indeed
many different factors can contribute to increasing the variability of test results such as the time interval between
measurements, the operator, the device used, the environment, and lead to the estimation of a
intermediate precision value (see standard NF ISO 5725-3).
Another equally essential quality of the test method is representativeness. It reflects the good correlation of the test
results with those carried out in service, on a real scale (for example: braking tests
on roller benches or on tracks).
We cannot design tests without measurement. Indeed, to assess the acceptable or unacceptable performance of a product,
it is necessary to quantify it:
— by a binary response (tire burst test, etc.);
— by a number of occurrences (number of cycles before failure in a fatigue test, counting in micro-
biology, etc.);
— by measurements of quantities (braking distance of a vehicle, sulfur content of diesel, etc.).
Furthermore, parameters specific to the test conditions (environmental conditions: humidity, ambient temperature, etc.)
must respect tolerances to comply with the test method. They are the
more often verified by measurements.
The result can, in certain cases, take on a conventional character, such as the determination of the
flash point of an oil cut (NF EN 22719) or the determination of the compressive strength of
concrete specimens (ISO 4012).
The measurement or testing process, without leading to a conventional result, may also not be modeled or modelable.
FD X 07-021 — 16 —
In all these cases, a more global approach to estimating the uncertainty associated with the measurement must be undertaken by
implementing cooperative tests called “interlaboratory tests” (NF ISO 5725).
The preliminary question for this evaluation is: “Is there a formalized and consensual method?”
If the answer is yes (possibly after drafting of a common operating procedure by a committee of experts),
Interlaboratory tests can be undertaken, following the requirements of standard NF ISO 5725. From this exercise, repeatability and
reproducibility can be evaluated which define the fidelity of the test method. When this evaluation is carried out as part of normative work,
these values can be retained as
precision values associated with the standardized method currently being developed.
If the formalized method is only used by a single laboratory, it can evaluate its own reliability; he must
document all the elements retained for the estimation of the uncertainty associated with the result.
If the laboratory participated in the interlaboratory test campaign and if its results are not significantly different from those announced by the
other participants or, in other words, if it was not rejected as
“aberrant” laboratory according to the recommended criteria (see NF ISO 5725-2), the laboratory can claim
the use of the precision values determined within the framework of the “interlaboratory test” to estimate its
uncertainty.
If the laboratory did not participate in the interlaboratory test campaign, it can estimate the uncertainty of its result by referring to the precision
of a standardized test method, provided that it provides proof that he uses the same test method and that he masters his measurement and
testing process. For this we can usefully
refer to the NF ISO 10012-2 standard “Quality assurance of measuring equipment — Guidelines for the control of measuring processes”.
5.7 Phase C4: Experimentation — Validation

5.7.1 Repetition of measurements and tests
Repeating measurements and tests is only of interest if the information provided by each observation is new,
and for this the experimenter must make the measurements and tests as independent as possible (not correlated).
The measurement or test method must be perfectly applied and carried out in such a way that there is no
influence on the consecutive results during repetitions (for example: vehicle presented with conditioning before each braking test; remove and
replace the probe between each opacity measurement test when this is imposed, useful or necessary, etc.).
The ability to perform an appropriate number of repetitions depends on the availability of resources and most
often the cost of the test or measurement.
The repetitions make it possible to estimate the repeatability, the value of which provides information on the possible existence of factors
important which would not have been taken into account until now in the modeling of the process of
measure.
5.7.2 Independence of repetitions of measurements and test results
The independence (non-correlation) of the repetitions of measurements and test results must on the one hand be ensured a priori during the
development of the measurement or test method and the organization of the experiments ,
on the other hand be verified a posteriori by means of appropriate statistical tests.
A priori, maximum technical precautions must be taken when developing the method of
measurement or test to ensure that during repetitions, the result of a measurement or test is not
influenced by measurements or test results that preceded it.
All useful details must be provided to the operating mode, to the preparation of the material, during the
carrying out consecutive measurements or tests.
For example, if zeroing a device is part of the operating mode, zeroing the instrument must
be part of every rehearsal. Another example, it is also necessary to ensure that the glassware is cleaned to eliminate any trace of product
from the previous test which could interfere with the following test.
During the experiment, the measurand depends not only on studied, controlled input quantities but also on variations in factors not
necessarily controlled (for example, wind conditions
on site, etc.). During the carrying out of the tests, these factors, which are difficult to control, are all the more unstable as the duration of the
tests is long.
— 17 — FD X 07-021
Experimental design techniques, including randomization, can be implemented in order to organize experiments to limit
and distribute these risks of influence as best as possible.
Repetitions of measurements or test results are most often organized in an order which requires as little manipulation
as possible for reasons of time and cost (change of operators, instrumentation, assembly and disassembly of
equipment). a device, etc.).
In this case, it is up to the person responsible for the experiment to find a balance between the requirements of scientific
rigor and economic constraints.
A posteriori, on the basis of the measurements or test results, it is possible to implement statistical tests to highlight a
possible dependence between the data.
These tests are easily accessible in the literature as well as the corresponding statistical software (for example the chi-
square test).
The limited scope of this approach should be emphasized in that it makes it possible to reject with a given level of
confidence the independence of the repetitions but that it cannot make it possible to remedy this dependence.
The implementation of independence tests cannot therefore replace the technical and organizational precautions to be
taken a priori.
5.7.3 Critical and physical examination of outliers
What do we call, or how can we define an outlier in a series of measurements?

Outliers are exceptional values which deviate so much from comparable values, obtained during the same test, that
they are considered incompatible with the latter. Such a definition allows the use of statistical tests to detect outliers.
What decision to make regarding an outlier?

The decision to accept or reject such a value is only made if the cause of the anomaly can be found; it could be a
technical error, for example a poorly executed test, or a calculation error, or a transcription error, etc.
If the cause of the anomaly has been detected and it can be corrected, the result is corrected as far as possible (e.g.
transcription error). Otherwise, the test or measurement is repeated.
If the anomaly has been detected but cannot be corrected, or if its cause could not be found, the elimination of the aberrant result
is the responsibility of the person responsible for the test (or the metrologist) and not of the statistician or the computer scientist.
The decision to reject a value with a low probability of occurrence, detected by a statistical test, is always delicate. For
example, the systematic elimination of extreme values (which are part of the distribution) from a series of results leads
to the underestimation of their variance. Likewise, in the case of very repeatable values, a result which deviates slightly
from the others may be rejected in the strict application of a test, even though this deviation is not physically significant.
Whatever the reason which leads to qualifying a measurement or test result as aberrant, the decision to accept or reject
it must only be taken with full knowledge of the facts (after reflection, analysis, investigation). , etc.) by the person best
suited to make the decision (physicist, metrologist, test manager, etc.).
5.8 Phase D1: Processing of measurements
5.8.1 Calculations of statistical parameters
Measurements concerned: the input quantities X1, X2, …, XN, that is to say the quantities which are used in the functional relation Y
= f(X1, X2, …, XN) and which are objects of measurement.
FD X 07-021 — 18 —
5.8.1.1 Calculations
Each input quantity can be estimated from a different number of measurements.
— the size of the sample is the number of measurements of the quantity Xi , xij being the jth measurement of the ith quantity,
with 1 ÿ j ÿ ni ;
— an estimate of the mathematical expectation of the distribution of the quantity from which the sample is drawn is the
arithmetic average
neither
1
----
xi = ÿ xij
neither
j=1
— an estimate of the standard deviation ri of the distribution of the quantity from which the sample is drawn is given:
- either by the experimental standard deviation:
neither
1 2
------------- %&
if = ÿ nor – 1 xij – xi
#$
j=1
from the sum of the squares of the differences between the values xij of the sample of the quantity Xi and the mean
xi values from this sample;
- or from the range, denoted wi , difference between the largest and smallest values.
For the sample of the quantity Xi , wi = max (xij) – min (xij), this value must be corrected to obtain an estimate
of the standard deviation of the magnitude distribution. To do this, and if we assume that the distribution law of
magnitude is normal, we divide the extent by a coefficient dni whose value is a function of ni , size of
the sample.
Table 2 which follows gives the values of the coefficient dni in the hypothesis of a normal law.
Table 2 — Value of dni as a function of ni
neither
2 3 4 5 6 7 8 9 10
dni
1.13 1.69 2.06 2.33 2.53 2.70 2.85 2.97 3.08
wi
------
.
The estimator of the standard deviation of the distribution is therefore:
dni
The reader's attention is drawn to the fact that these two estimates lead to different numerical values but considered equivalent
in this context.
NOTE If ni is large (greater than 10), the best estimator of ri is the experimental standard deviation. If ni is low (lower
to 10), we can use the range wi , which is simpler to calculate, by correcting the value by 1/ dni . In both cases, the
confidence interval of the estimator obtained ri can be evaluated using statistical tables (see specialized works).
5.8.2 Analysis and use of statistical parameters: calculation of y estimation of measurand Y
5.8.2.1 Analysis
The means and standard

xi deviations si are used to determine the estimate of the measurand Y represented by y f(x1, x2, …, xN)
= that the quantities Xi remain within limits of specification.
and/or to verify
EXAMPLE
We measure the temperature of a gauge block and verify that this temperature is within the limits 20 °C ± 2 °C
in order to validate the measurement. A temperature reading of 23°C would make it possible to determine the length of the hold (in
making the relative temperature correction to 23°C), but would show that the measuring bench was not used
in good conditions (temperature outside the specifications for use).
— 19 — FD X 07-021
Estimates of standard deviations are used to estimate the standard uncertainty associated with each quantity, but are also
used:
— to validate the measurement and assessment of uncertainty (e.g. verification that repeatability is
sufficient);
— to verify a posteriori the hypotheses made in the process:
- is the hypothesis of small variations around the average value justified? - are the various
contributions of the input quantities to the uncertainty of the same order of magnitude? - does the model chosen to
model the measurement process adequately account for the measurements taken, taking into account the uncertainty
at each point? If this is not the case, a fundamental technique is the examination of residues (see specialized
works). Otherwise, the model must be modified.
5.8.2.2 Use statistical parameters of
Example of sensor calibration, model verification.

It is assumed that the sensor response follows the linear law Y = f(X). For example Y is a tension, X is a force
generated by a standard.
For a set of values x describing the range of use of the sensor, we calculate the corresponding y from the model y = ax +
b. The comparison of the points calculated according to the model and the experimental points, taking into account the
uncertainties on the x and the uncertainties on the y, makes it possible to determine if the chosen model is satisfactory,
, the model line is attributable or not to measurement hazards.
that is to say if the dispersion of the points {x y} around
Otherwise, the model needs to be revised, for example replaced by y = a x2 + bx + c to account for a non-linear response.
5.9 Phase D2: Law of propagation of measurement uncertainty

After estimating the standard uncertainties on each of the model input quantities, the uncertainty propagation law makes it
possible to “propagate” the uncertainties of the different quantities to calculate the composite standard uncertainty uc(y)
on the measurement result.
5.9.1 Composite standard uncertainty

It is written in its general form for correlated input quantities 1) :
NOT 2 N–1 NOT
2 = ÿf ÿf ÿf
uc ( ) y ÿ -------
ÿxi
2u
xi ( ) 2
+ÿ ÿ -------
-------u xi ,xj
( ) ÿxi
... (1)
i=1 i=1 ÿxj ji = + 1
The covariance term can also be written:
u(xi , xj ) = r(xi , xj ) ÿ u(xi ) ÿ u(xj )
with: – 1 ÿ r(xi , xj ) ÿ + 1
r(xi , xj ) being an estimate of the linear correlation coefficient between Xi and Xj .
This expression can be simplified depending on whether the input quantities Xi are all correlated in pairs or not.
5.9.2 Correlation
There is correlation (covariance) when a common doubt weighs on two estimates xi and xj of the variables Xi and Yj which enter into
the composition of the result y Seeking to know whether two input
.
quantities Xi and You should not reason physically by thinking about the two corresponding measurands. On the other hand, it is necessary to
analyze how the measurement process could introduce, or could not eliminate, a doubt common to these two random variables which represent
the measurands.
1) See 13 5.2.2
the equation
of NF ENV 13005. in standard
FD X 07-021 — 20 —
Two important special cases — The input
quantities are uncorrelated.
Which means :
NOT 2
2 % &* ÿxi
ÿf )
-------
2u
whatever i and j, r(xi , xj ) = 0, then, uc ( ) y #$ xi ( )
=ÿ
i=1
Then the square of the standard uncertainty composed of y is written as the sum of the squares of the standard uncertainties
%-------
&
ÿf
of xi weighted by their sensitivity coefficient ) * ÿxi # $ squared.
%-------
& ÿf
By asking : ) * = Ci
# ÿxi
$
2 22u 2 2 2
=
uc ( ) y C1 x1 ( ) C2 x2u()… CN + ++ ( ) xN 2 u ... (2)
— The input quantities are all perfectly positively correlated two by two.
2
NOT
%-------
&
xi ÿfu
ÿ
2
Whatever i and j, r(xi , xj ) = 1, then, units ()y = ( ) ) * ÿxi # $
i=1
The square of the composite standard uncertainty of y is written as the square of the sum of the standard uncertainties of xi
%-------
&
ÿf
weighted by their sensitivity coefficient ) * ÿxi # $
%-------
& ÿf
By asking : ) * = Ci
# ÿxi
$
2
2
= ... (3)
units ( ) y C1u x1 ( ) C2u x2 ()… CNu xN + ++ ( )
5.10 Phase D3: Use of loyalty values
A laboratory provides a result to which it must associate an uncertainty.
The uncertainty of the test result must reflect the test conditions, means and resources implemented by the laboratory in which the test was carried out. All the
test conditions, means and resources implemented by the laboratory are an integral part of the test method used, the variability of which is quantified by the
precision (NF ISO 5725).
Precision can take two extreme values: repeatability and reproducibility defined in the conditions of implementation of the method.
Repeatability conditions : Conditions where independent test results are obtained by the same method on identical test individuals in the same laboratory, by
the same operator using the same equipment and during a short interval of time (NF ISO 3534- 1).
Reproducibility conditions : Conditions where test results are obtained by the same method on identical test individuals in different laboratories, with different
operators and using different equipment.
However, in a laboratory, repeatability conditions are not always strictly respected (tests at different times, calibration conditions between tests, different
operators, different equipment, etc.). These situations give rise to the estimation of intermediate fidelities in a given laboratory using appropriate statistical
models (NF ISO 5725-3).
— 21 — FD X 07-021
In what follows, the term intermediate fidelity is used in the broad sense and includes the notion of repeatability.
In practice, we use intermediate precision, the conditions of strict repeatability being rarely satisfied, we
therefore endeavors to define conditions of intermediate precision corresponding to the current use of the method
(e.g. operator factor, time factor, etc.).
5.10.1 The laboratory participated in an interlaboratory test
5.10.1.1 Reproducibility of the method
In many practical cases we have to use the standard deviation of reproducibility sR determined during
the interlaboratory test, as an estimate of the standard uncertainty (see NF ISO 5725-3).
uc(y) = sR
This solution has advantages (simplicity), but also disadvantages (standardization of the uncertainty for all laboratories, sR is certainly an
increase in the standard uncertainty and the use of sR leads
implicitly to a change in the measurand, we are no longer interested in the value announced by a laboratory,
but at the value that could have been announced by a set of laboratories). Laboratories that wish
express their uncertainty more finely are invited to use the methods proposed according to 5.10.1.2 and 5.10.1.3.
It is therefore necessary to justify this uncertainty.
5.10.1.2 Common intermediate loyalty
Testing laboratories must be able to use the common value of the intermediate precision standard deviation sI
.
as an estimate of the composite standard uncertainty uc On the other hand, the method of estimating the intermediate precision ensures
that the results of the different participants are sufficiently close, so that a laboratory presenting a precision greater than that common , is
not penalized and can thus use common loyalty.
uc(y) = sI
5.10.1.3 Laboratory-specific intermediate precisionof
A laboratory with better fidelity must be able to use its own sIo value provided that it can
prove that he has carried out additional tests or used another statistical model allowing him to quantify the sources of uncertainty more
precisely.
uc(y) = sIo
5.10.2 The laboratory did not participate in an interlaboratory test
5.10.2.1 Published reproducibility of theof

method
In many practical cases we have to use the published reproducibility standard deviation sR as an estimate of the standard uncertainty.
uc(y) = sR
This solution has advantages (simplicity), but also disadvantages (standardization of the uncertainty for all laboratories, sR is certainly an
upper limit of the standard uncertainty, and the use of sR leads
implicitly to a change in the measurand, we are no longer interested in the value announced by a laboratory,
but at the value that could have been announced by a set of laboratories). Laboratories that wish
express their uncertainty more precisely are invited to use the methods proposed according to 5.10.2.2 and 5.10.2.3.
It is therefore necessary to justify this uncertainty.
NOTE 1 Standard NF ISO 5725-3 deals with the possible relationship of the standard deviation of reproducibility with the level
(concentration, value of the measurand).
NOTE 2 For agro-food analyses, there is experimental work which has led to the development of a
conventional equation, expressing the relative standard deviation of reproducibility (in percent) as a function of the concentration C
expressed as a percentage:
sR = 2(1 – 0.5 log10C) or sR = 2C – 0.1505

FD X 07-021 — 22 —
5.10.2.2 Intermediate precision published in the method
In the event that a laboratory did not participate in the interlaboratory test and wishes to use as an estimate of
the standard uncertainty uc the published value of sI , it is essential that he can prove that he is using the correct method
test as well as the same statistical model for analyzing the results as that implemented in the interlaboratory test and that its
intermediate precision is not significantly different from the published precision.
uc(y) = sI
5.10.2.3 Laboratory specific intermediate precisionof
The laboratory is entitled to estimate the standard uncertainty uc using an intermediate precision determined on its own test
results, provided that it provides proof of mastery of the method of testing and implementation
implementation of an appropriate statistical model.
uc(y) = sIo
5.11 Phase D4: Expression of the result and its uncertainty
The application of phases D2 or D3 makes it possible to estimate a composite standard uncertainty uc(y). For different reasons
we may be led to express an expanded uncertainty U, such that:
U = k ÿ uc(y)
Or :
k is the broadening factor whose value is chosen on the basis of the level of confidence required for
–
the interval [y U, y + U]. In general k is equal to 2 or 3.
Choosing k = 2 conventionally amounts to considering an interval with a confidence level of approximately 95%.
Choosing k = 3 conventionally amounts to considering an interval with a confidence level of approximately 99%.
The numerical values of the estimate y and its standard uncertainty uc(y) or of U must not be given with
an excessive number of digits. It is usually sufficient to provide uc(y) and U to two significant figures.
The result of the measurement or test is then expressed in the form:
Y = y ± U for a given enlargement factor.

An estimate of a measurement or test result must be rounded consistent with its uncertainty. For example if y = 10.057 62 ÿ with uc(y) = 27
mÿ, y should be rounded to 10.058 ÿ.
5.12 Phase E1: Decision
From the available data which are:

— the measurement or test result and its associated uncertainty;
— a judgment limit (specification limit, class limit, control or monitoring limit),
decisions can be made.
Figure 3 illustrates a decision strategy.

— 23 — FD X 07-021
Figure 3 — Illustration of a decision strategy
NOTE 1 These rules apply to a corrected measurement result (correction of presumed systematic errors). Further normative
work should complete this issue of use of measurement uncertainty.
NOTE 2 In the case where the test involves sampling, if the definition of the measurand does not include the representativeness
of the sample with respect to the lot, then the conclusions are only applicable to the sample and to only him.
6 Appendices
Appendices A to H present examples of estimating measurement uncertainties. These examples use the procedure
described in this document. They were created for the purposes of illustrating the approach proposed for estimating
uncertainties; the numerical values indicated are not necessarily actual data.
FD X 07-021 — 24 —
Annex A
(informative)
Estimating the calibration uncertainty of a liquid expansion thermometer
Init numbering of annex tables [A]!!!

Init numbering of appendix figures [A]!!!
Init numbering of appendix equations [D]!!!
This example illustrates a simple case of estimating measurement uncertainty.
A.1 Phase A1: Definition of the need

Calibration of a liquid (mercury) expansion thermometer: —
Measuring range: 40°C to 65°C;
— Graduation interval: 0.1°C.

This thermometer is used to make temperature measurements with maximum permissible errors of +0.2°C and –0.2°C,
only at calibrated temperature points.
A.2 Phase B1: Means used

The calibration chain consists of: — a 100
ÿ platinum resistance thermometer;
— a multimeter with a resolution of 0.1 mÿ;
— a transfer bath enabling temperatures between – 20°C and + 85°C to be obtained.
The expanded uncertainty (k = 2) of this reference chain is U = 0.06 °C.
A.3 Phase C1: Analysis of the measurement process

The causes of errors taken into account are shown in Figure A.1.
Figure A.1 — Error cause diagram

— 25 — FD X 07-021
A.4 Phase C2: Modeling the measurement process

We are in the case of a direct measurement; there are therefore no intermediate quantities intervening in a functional relationship.
The model input quantities are the reading and the corrections, i.e.:
h = +hCe + Cq + Cc
A.5 Phase C4: Measurements carried out
Measurements carried out at ambient temperature ha of 20°C ± 1°C, at 5 points of the range (40°C ÿ 65°C),
are repeated twice.
The height of the emerging column is constant, whatever the temperature read and is represented by 1 °C
on the thermometer scale. The values read are given in Table A.1.
Table A.1 — Statement of values read
Values read
Reference (°C)
(°C)
Test 1 Test 2
40 40.00 39.98
45 45.00 45.00
50 50.02 50.02
55 54.98 55.00
60 60.00 60.02
NOTE The reading, made with a cathetometer, allows 1/5th of the graduation to be appreciated.
A.6 Phase D1: Estimation of standard uncertainties

Repetition of observations: it is estimated from the maximum extent (Wmax) of the values read, by
Wmax
------------
the expression .
dn
Table A.2 gives values of dn as a function of n gives dn = 1.13 for n = 2.
Table A.2 — Values of dn as a function of n
not 2 3 4 5 6 7 8 9 10
dn 1.13 1.69 2.06 2.33 2.53 2.70 2.85 2.97 3.08
The standard uncertainty associated with the repetition of observations is therefore:
2
= ----------- × 10-2 = 1.18 × 10-2 °C
uh
1.13
— Calibration bench: the expanded uncertainty (k = 2) of the calibration chain is U = 0.06 °C; it corresponds to
0.06
-----------
= 3 × 10-2 °C.
the standard uncertainty uce = 2
FD X 07-021 — 26 —
— Reading: the reading, made with a cathetometer, allows one to appreciate 1/5 of the graduation. The resolution is as follows:
0.06
-----------
r= × 10-2 = 2 × 10-2 °C
2
The value of the corresponding standard uncertainty is obtained by assuming a rectangular law, i.e.:
0.06
-----------
× 10-2 = 0.6 × 10-2 °C
ucq = 2
— Emerging column
The error due to the emerging column is obtained by applying the following correction law:
hc = h1 + A ÿ n(h1 – ha)
with :
h1 = temperature read on the thermometer;

Ha = ambient temperature around the emerging column;
not
= emerging column, expressed in number of degrees Celsius;
A = 0.16 × 10-3 = apparent cubic expansion coefficient of mercury in glass.
This error is maximum at the measurement point h = 60 °C and ha = 19 °C; it is equal to:
ÿh = 0.16 × 10-3 × 1(60 – 19) = 0.6 × 10-2 °C
This error is not corrected (Cc = 0 in the model). It is taken into account in the form of a
standard uncertainty by assuming a rectangular distribution law:
0.6
----------
ucc = 2 3 × 10-2 = 0.2 × 10-2 °C
NOTE It may happen that this error, uncorrected (and not considered in the uncertainty balance), is combined linearly with the
expanded measurement uncertainty. This way of doing things is envisaged in standard NF ENV 13005 (article 6 and
annex F) with great caution and in very special circumstances; it should therefore not be recommended.
The standard uncertainties are summarized in Table A.3.
Table A.3 — Summary of standard uncertainties
Standard uncertainty
Uncertainty component Estimation method in 10-2 °C
Repeatability Extent 1.8
Calibration bench Calibration certificate 3.0
Reading Rectangular law 0.6
Emerging column Rectangular law 0.2
A.7 Phase D2: Law of propagation of measurement uncertainty
The causes of uncertainty being, obviously, statistically independent, the law of propagation of uncertainty is:
2 2 2 2 2 2 2 2 2
units = +++
C1 uh C2 uce C3 ucq C4 ucc
like C1 = C2 = C3 = C4 = 1
— 27 — FD X 07-021
ÿh ÿh ÿh ÿh
since ------- ===-------
------- -------- =1
ÿh ÿthis ÿcq ÿcc
2
uc = =
yes × 10-2 °C
ÿ 1.82 32 0.62 0.22 + +
uc = 3.6 × 10-2 °C
Or an expanded uncertainty (k = 2): U = 7.2 × 10-2 °C.
A.8 Phase D4: Expression of the measurement result

h = reading ± 7.2 × 10-2 °C (k = 2)
A.9 Phase E1: Decision
The comparison of the expanded uncertainty U to the maximum tolerated errors (± 0.2 °C) can only be made if the
conditions of use do not lead to a significant degradation of uncertainty. Otherwise, it is
the uncertainty obtained by taking into account the conditions of use which must be compared to the maximum tolerated
errors, in order to make a decision as to the suitability of the thermometer.
FD X 07-021 — 28 —
Appendix B
(informative)
Estimation of the uncertainty of a measurement of polluting CO emissions in the automobile
industry
Init numbering of annex tables [B]!!!
Init numbering of appendix figures [B]!!!
Init numbering of appendix equations [E]!!!
For this example, the illustration of the flowchart (Figure 2) is given in Figure B.1 in which the lines in
bold represent the path specific to this uncertainty calculation.
NF ISO 5725 NF ENV 13005
Figure B.1 — Application of the flowchart to CO measurement

— 29 — FD X 07-021
NOTE The bold lines represent the path specific to this uncertainty calculation.
Figure B.1 — Application of the flowchart to CO measurement (end)

FD X 07-021 - 30 -
Appendix C
(informative)
Estimating the uncertainty associated with measuring the height of the first bounce of a tennis ball
Init numbering of annex tables [C]!!!

Init numbering of appendix figures [C]!!!
Init numbering of appendix equations [F]!!!
This example illustrates the benefit of estimating uncertainty to determine the useful number of repetitions. Furthermore, it takes into account
correlation problems.
C.1 Phase A1: Definition of the need
This test consists of measuring the height of the first bounce of a tennis ball dropped from a defined height.
In order to define the operating mode, we seek to determine the number of repetitions necessary to obtain an uncertainty associated with the
average of the rebounds of the order of a centimeter. To do this, we draw the curve describing the uncertainty as a function of the number of
repetitions.
C.2 Phase B1: Choice of resources

Definition of bounce height: distance from the ground to the bottom of the ball.
Description of the equipment and measurement (Figure C.1): The ball is
released from the release arm with zero speed. After reaching the ground, it bounces back. The passage of the ball in front of the lower cell
then triggers a timer which is then stopped when the ball passes the upper cell. This recorded duration (ÿt), the acceleration of gravity (g) and
the distance between the two cells (d) make it possible to calculate the height of the end of the ball's trajectory (h).
By adding the distance from the ground of the bottom cell (A) and subtracting the diameter of the ball (D) because the cells fire at the top of
the ball, we can calculate the bounce height (r).
Legend Ratings
1 Release arm A Distance from the ground to the lower cell
2 Ball
(in m) d Distance between the two cells (in m)
3 High cell
g Acceleration in the test room (in m/s2)
4 Low cell
ÿt Duration recorded between the passage of the ball in front of the two cells (in s)
5 Ground D Diameter of the ball considered (in m) r Rebound
height (in m)
Figure C.1 — Machine diagram

— 31 — FD X 07-021
C.3 Phase C2: Process modeling

2
2d
------
2 – ÿt
g
--------------------
r AdhD = + + – with h = f1(d, g, ÿt)=
2ÿt
therefore r = f2(A, d, g, ÿt, D)
2 2
d d gt ÿÿ – D
r=A+d+ ------------------- – +-----------------
22
8 2gÿÿ
t
2
d
-- d 2gt ÿÿ
------------------ +--------------- – D
r = A ++
2 28
2g t ÿÿ
r
The result for which we seek uncertainty is the average of n ball bounce heights.
The process model is as follows:
not not
2 ÿ
2
1 1 d d g ti ÿ
r = --
not
ÿ laughed
-- HAS
= ÿ not
+ ++ ------------------
--------------------
2 2 8
–D
i=1 i=1 2g tiÿÿ
C.4 Phase D2: Application of the law of propagation of uncertainty
not
1
--
This is the law of propagation of uncertainty to the function r laughed
.
= ÿ not
i=1
% &
not n–1 not
2 1
uc ( ) r
= -----
2
not
ÿ 2 ÿ+ 2u
laughed ( )
ÿ ri rj
u( ), - we must not forget the covariances -
i=1 i = 1 ji = + 1 *
)# $
2
u laughed ( ) = u(r) whatever i
0
hypotheses 1
0 u (ri)rj=, / cov whatever i and j
It is assumed that the standard uncertainty associated with the rebound height is constant (for all repetitions), as is the covariance between two
repetitions.
The equation then becomes:
2 1 2 2 1
= ----- naked ÿ cov = 2 -- u ( )r + ( ) n – 1 ÿ cov
( )r 2+ Cn
ÿ ÿ
uc ( ) r 2 not
not
the different terms of which are calculated as follows:

2
— Calculation of Cn
2 2 not! nn ÿ (2 ) – 1
= ------------------------ = ------------------------
Cn is a combination: Cn
2!( ) n – 2 !
FD X 07-021 — 32 —
— Calculation of u2(r)
Or :
2
d
-- d 2gt ÿÿ
------------------ +---------------
r = A ++
2 28
2g t ÿÿ
The input quantities A, d, g, D, ÿt being assumed to be independent, there are no covariance terms in
the estimate of u2(r), therefore:

2 2 2 2 2
2 ÿr
------
2 ÿr
-----
2 ÿr
-----
2 ÿr
-------
2 ÿr
--------
2
uc ( )r
= ÿÿÿ ÿ u ( ) HAS + u ()d+ u ()g+ u ()D+ ÿ
u ( ) ÿt
ÿA ÿd ÿg ÿD ÿÿt
— Calculation of u(ri , rj ) (noted cov)

The distances (A and d), the diameter of the ball (D), the acceleration of gravity (g) are not remeasured at
each trial, they introduce a covariance between the bounce height measurements.
2 2 2 2
ÿr
------
2 ÿr
-----
2 ÿr
-----
2 ÿr
-------
cov = ÿÿÿ ÿ u ( ) HAS + u ()d+ u ()g+ 2 u () D
ÿA ÿd ÿg ÿD
The final formula for the variance of the mean is:
2 2 2 2 2
2 ÿr 2 ÿr 2 ÿr 2 ÿr 2 1 ÿr
uc ( ) r
= ÿÿÿ ÿ
------
u ( ) HAS + -----
u ()d+ -----
u ()g+ -------
u () D + -- -------- ÿ
2u( ) ÿt
ÿA ÿd ÿg ÿD not ÿÿt
Either by developing:
2 2
2 2 1 d 2 – d2 2ÿt 2 2 1
-- – d2 gt ÿ ÿ
-- --------------------- ---------------
uc ( ) r
= ÿÿ ÿ u ( ) HAS + +--------------- u ()d+
2
+ ------- u ( ) gu+ () D+ +------------ 2u( ) ÿt
2 8 n 4
gt 2 ÿ ÿ 2g 2 ÿ ÿt gt 3 ÿ ÿ
The estimate of the standard uncertainty of each component is given in Table C.1.
Table C.1 — Estimated standard uncertainty of each component
Variable Value Instrument Source Fashion ÿr

-----
Unit u(x) C= 3C ÿ u(x)3
X x measuring uncertainty estimation ÿx
HAS 1.105 m ruler resolution: 0.001 Rectangular law 2.9 × 10-4 1 2.9 × 10-4
metallic 1
d 0.1 m ruler resolution: 0.000 5 Rectangular law 1.44 × 10-4 7,000 9 1 × 10-3
metallic 2
g 9.809 22 m/s2 gravimeter calibration Certificate 4.583 × 10-3 – 3.33 × 10-2 1.53 × 10-4
calibration
D 0.065 1 m gauge limits: Rectangular law 9.24 × 10-4 –1 9.24 × 10-4

[0.063 5; 0.066 7]
ÿt 0.039 6 a) s stopwatch Standard deviation 7.7 × 10-4 – 16.32 1.257 × 10-2

(Resolution : experimental
0.000 01)
(2.9 × 106)
(comments
repeated
measurements) b) (7.7 × 10-4)
has) Average passage time on b) 8 balls × 5 repetitions.

Standard uncertainty estimated with measurements accumulated during time.
— 33 — FD X 07-021
C.5 Phase D4: Expression of the result and its uncertainty

Table C.2 gives the expanded uncertainty as a function of the number of repetitions (n).
Table C.2 — Expanded uncertainty

as a function of the number of repetitions
Rehearsals Expanded uncertainty (in m)

(not) (k = 2)
1 0.025
2 0.018
3 0.015
4 0.013
5 0.012
6 0.011
7 0.009 9
8 0.009 3
9 0.008 8
10 0.008 4
We can also present the expanded uncertainty as a function of the number of repetitions in the graphical form of the
Figure C.2.
0.04
0.03
1
0.02
0.01
0.00
1234 5 6 7 8 9 10
Legend
1 Expanded uncertainty (m)
2 Number of repetitions
Figure C.2 — Expanded uncertainty as a function of the number of repetitions
C.6 Phase E1: Decision

After calculating the expanded uncertainty of the average of n bounce heights, it was decided to use this
equipment by making seven repetitions for each test, which allows the result to be announced to within ± 1 cm.
FD X 07-021 — 34 —
Appendix D
(informative)
Estimation of the uncertainty of a current intensity measurement
Init numbering of annex tables [D]!!!

Init numbering of appendix figures [D]!!!
Init numbering of appendix equations [G]!!!
This example illustrates a case of improving the quality of the measurement thanks to measurement uncertainty.
D.1 Phase A1: Definition of the need
Consider an electrical circuit in which a direct current of approximately 10 A circulates. A simple and easy-to-implement
means is required for measuring this current. The context may be that of school practical work.
Not giving any indication on the circuit, the measurand must be defined as being the intensity of the current in the presence
of the measuring system, so as not to have to take into account the disturbance brought by this measuring system.
D.2 Phase B1: Choice of resources
The measurement principle, which implements Ohm's law, is illustrated in Figure D.1.
Figure D.1 — Principle of measurement
The equipment consists of a shunt and a digital voltmeter. The current intensity is close to 10 A, the shunt has a nominal
value of 0.01 ÿ. The voltmeter is assumed to be almost ideal, with a negligible measurement current and an input impedance
greater than 10 Mÿ.
The operator makes the connections and reads the value indicated on the voltmeter. He repeats this operation 12 times.
The ambient temperature is 23°C ± 3°C.
A priori, within the limits of the proposed point of view, errors occur due to:
— readings on the display which are not stable;
— the shunt which has been imperfectly calibrated, at a temperature which is not that at which it is used; — the
voltmeter for which the manufacturer's specifications have been verified.

— 35 — FD X 07-021
In the context of a metrology laboratory, we would have to take into account other phenomena
such as :
— the effect of inserting the measurement system;
— parasitic electromotive forces of thermal origin;
— the power coefficient of the shunt;

- etc.
D.3 Phase C1: Complete analysis of the process
The value obtained results from the application of Ohm's law. The input quantities (voltage U and resistance R) are
or should be corrected by:
CV : Correction to the voltage read on the voltmeter. This correction is not evaluated; we assign a value to it
zero, by associating it with a standard uncertainty deduced from the manufacturer's verified specifications;
CE : Correction on the resistance taken from the shunt calibration certificate. The associated standard uncertainty is deduced
the calibration certificate;
Ch : Correction on the resistance, of the influence of the temperature different from that during calibration. This
correction is not applied: it is assigned a zero value, associating it with a deduced standard uncertainty
environmental conditions.
Readings fluctuate randomly. Or the mean of the sample, to which is associated a standard uncertainty determined by a type A
evaluation method. The values read on the voltmeter being close, the correction that should be made to them is assumed to be
independent of these readings.
D.4 Phase C2: Process modeling
The process can be modeled by applying Ohm's law and we therefore obtain:
U
= = --- CV +
I --------------------------------
R + +Ch
RE CE
Or :
RE designates the nominal value of the shunt.
D.5 Phase D1: Processing of measurements
D.5.1 Processing of readings
The n readings are processed using a type A evaluation method. We have:
not
1
--
i
= ÿnot
i=1
and the corresponding standard uncertainty given by the experimental standard deviation of the mean:
not
2
1
u( ) = s( ) = --------------------
n n( ) – 1 ÿ % i– #
&
$
i=1
The numerical calculation is presented in Table D.1.

FD X 07-021 — 36 —
Table D.1 — Statement of values read
Reading no. i
(mV)
1 100.13
2 99.98
3 99.94
4 100.09
5 100.2
6 99.93
7 99.98
8 99.9
9 100.06
10 100.15
11 100.06
12 99.94
Average 100.03
Experimental standard deviation 0.098 5
Experimental standard 0.028 4

deviation of the mean
D.5.2 The voltmeter
We make the simplifying assumption that the voltmeter is not calibrated and that we rely on the specifications of the
constructor.
It is used on the 200 mV gauge for which it is specified that the maximum error, for use at an ambient temperature between 15°C
and 30°C, is the sum of 0.05% of the voltage read and 0.02% of the caliber
used, i.e. a maximum error bV = 90 µV.
This specification has been verified; the error is assumed to be equiprobable within the interval [ÿ 90 µV, 90 µV].
The maximum error is divided by 3 1.7 ÿ to obtain the standard uncertainty:
u(CV) = 52 µV
On the other hand, we have obviously neglected the effects of the input impedance of the voltmeter, which is very large compared to
to the shunt, and a possible input current, negligible compared to that measured.
D.5.3 The shunt
The shunt has been calibrated under its conditions of use, namely for an ambient temperature of 23°C and under
a current of 10 A. The result is:
R = 0.010 018 ÿ with a relative calibration uncertainty of 6 × 10-4 (k = 2).

— 37 — FD X 07-021
We can deduce :
u (CE) = 3 µÿ
Shunt temperature is not measured. We are therefore not able to make any corrections. The maximum error can be
deduced from the tolerance on the ambient temperature which is ÿt = 3 °C, the temperature coefficient
ah between 15 °C and 28 °C being in relative value of 5 × 10-5 K-1. This leads to a maximum error in value
absolute given by:
ah ÿ ÿt ÿ R = 1.5 µÿ
We make the pessimistic hypothesis that the ambient temperature oscillates between the limits. The estimation of the standard
2 1.4 ÿ .
uncertainty is therefore done by dividing by
u(Ch) = 1.06 µÿ
D.6 Phase D2: Uncertainty propagation law

The law of propagation of uncertainty leads to:
2 1 U2
= ------ u2 ( ) u 2 2
uc ( )I ThisU+ ( ) Ch ]
2 ------ u
CV + ( ) + [
R2 R4
It could also be written in relative value:

2
uc ( )I 1
------------ = ------ u
2 2 1 2
( )u
Resume + ( ) + [
2 ------ u
ThisU+ ( ) Ch ]
2
I U2 R2
D.7 Phase D4: Expression of the result and its uncertainty

The measured current is therefore I = 9.985 A.
Table D.2 presents a summary of the calculations.
Table D.2 — Summary of calculations
Sensitivity Term
Fashion Standard
Variable Uncertainty component of the sum (mA2)
estimation uncertainty
coefficient
if = u( ) Dispersion of readings Standard deviation 28.4 µV 8,091

experimental 1/R
102 ÿ-1
u1 = u(CV) Voltmeter specification Law 52µV 27
rectangular
u2 = u(CE) Shunt calibration Certificate 3 µÿ 9

calibration U/ R2
103 A.ÿ-1
u3 = u(Ch) Influence of temperature on the shunt U-shaped law 1.06 µÿ 1,125
Variance estimation 45.22mA2
uc Compound standard uncertainty 6.72mA
U Expanded uncertainty (k = 2) 13mA
The final result can be stated in the form:

I = (9.985 ± 0.013) A (k = 2)
FD X 07-021 — 38 —
D.8 Phase E1: Decision — Use of the result and its uncertainty
In the event that the uncertainty obtained is too great, Table D.2 helps guide modifications to the measurement system. It would be necessary
successively to:
— have the voltmeter calibrated;
— increase the number of readings or use a more stable generator;
— have the shunt calibrated with better uncertainty;
— measure the shunt temperature and apply a correction;
- etc.
D.8.1 Calibration of the voltmeter

A first possible improvement is the calibration of the voltmeter instead of checking the manufacturer's specifications. It is common practice for
the ratio between the verified tolerance bV and the extended calibration uncertainty of the metrological means implemented for this verification
to be at least equal to 4. The use of calibration leading to the observation verification would reduce this component to:
bV
----- ÿ --
1
u(CV) = 4 k
The new value of this standard uncertainty is:
u(CV) = 11.25 µV
D.8.2 Optimization of the number of readings

It is possible to reduce the influence of random errors by increasing the number of readings. On the other hand, there is no point in increasing
it recklessly.
The composite standard uncertainty can be expressed by the equation:
2 sL
uc = + ----2 one
In this equation, sL designates the experimental standard deviation of the sample constituted by the n values read
sL
------
(the sensitivity coefficient being taken into account) and u the quadratic sum of the other components. If s = ,
not
we can agree to choose n such that s is equal to half of u, thus s intervenes for approximately a tenth in
2
%--
&
1) *
+ 1 ÿ u ÿ 1.12 u .
2
#$
sL
------
The curve in Figure D.2 represents the variations of uc, u and s = verified. , as a function of n, in the case of the voltmeter
not
— 39 — FD X 07-021
666 Composite standard uncertainty: uc

––––
Component estimated by a type A method: s
–– –– Component estimated by a type B method: u
Figure D.2 — Case of the verified voltmeter
2
u %sL&
Let us therefore solve for s = 2 -- , which leads to n = 4) * ----
. In the example above, we would get n from the order
u
#$
of 11. The results are summarized in Table D.1.
Table D.3 — Summary of results in the case of an audit
s Component estimated by a type A method 2.97mA
u Component estimated by a type B method 6.09mA
U Expanded uncertainty 14mA
If the voltmeter is calibrated, the curve becomes that of Figure D.3.
666 Compound standard uncertainty

––––
Component estimated by a type A method
–– –– Component estimated by a type B method
Figure D.3 — Case of the calibrated voltmeter

FD X 07-021 — 40 —
We then obtain n ÿ 34. The results are summarized in Table D.2.
Table D.4 — Summary of results in the case of calibration
s Component estimated by a type A method 1.69mA
u Component estimated by a type B method 3.375mA
U Expanded uncertainty 7.5mA
This progress in reducing uncertainty shows that it is certainly useful to include in the list
uncertainty components those whose participation in the value of the compound uncertainty is negligible.
This may initially be true, but should be reconsidered during the improvement process.
— 41 — FD X 07-021
Appendix E
(informative)
Estimation of the uncertainty of a measurement of the luminance coefficient of
pavement marking materials
Init numbering of annex tables [E]!!!

Init numbering of appendix figures [E]!!!
Init numbering of appendix equations [H]!!!
This example illustrates the application of the approach to road marking.
E.1 Phase A1: Definition of the need
The photometric characteristics of materials are important to know in the road sector (perception of the roadway by the
motorist) and in urban planning (perception of a complex environment
by a person located anywhere on the site). They determine the safety and comfort of users.
The reflection properties of a material are determined via the luminance coefficient q, quotient of the luminance L of a surface
(light impression received by an observer looking at the surface) by
the illumination E received by this surface. It is expressed in cd ÿ m-2 ÿ lx-1.
L
---
This coefficient q = E , is the measurand for which we want to know the measurement uncertainty.
For a given coating, it depends on two directions defined by four angles (see Figure E.1).
Legend
1 Light source
2 Point observed
3 Observer
Ratings
a Viewing angle
b Angle between the lighting plane and the observation plane
c Illumination angle
d Angle between the observation plane and the axis of the road
Figure E.1 — Definition of angles for measuring the luminance coefficient

FD X 07-021 — 42 —
E.2 Phase B1: Choice of resources

The gonioreflectometer (see Figure E.2) was developed to photometrically characterize a
material sample from the values of the luminance coefficients for different values of the angles
defined above (except for d, the material being assumed to be isotropic in luminance). A game of mirrors and
motors allows you to vary the angles and therefore explore a half-sphere above the sample (see the
Figure E.3). A diaphragm keeps the illuminated surface constant as the angle increases.
lighting increases.
Legend
1 Cell 5 Light source
2 Lighting arms 6 Sample

3 Measuring arm 7 Tray
4 Diaphragm MT Motor
Figure E.2 — Schematic diagram of the gonioreflectometer
Legend
1 Source
2 Observer
Figure E.3 — Monitoring zone of the device

— 43 — FD X 07-021
The principle of measurement is as follows:
The illuminance on the surface S (10 cm by 10 cm) is:
E = Eÿ ÿ cos c
Eÿ being the illuminance measured in the plane perpendicular to the incident direction of the light using a
light meter.
The luminance L of the surface in the measurement direction making the angle a relative to the surface is determined by measuring the light
intensity reflected by this surface:
------------------
L=
S ÿ sina
The source being assumed to be point, is determined by measuring the illumination Er received by the cell (luxmeter) from the sample located at
a distance d (150 cm) and by applying the relationship:
2
= Er ÿ d
The functional relationship expressing the physical law of the coefficient q (measurand) is:
2
Erd
= -----------------------------------
q
SEÿcoscsina
For a conductor located 1.5 m from the ground, the angle a is between 0.5° and 1.5°, and as in this range the variations in q are small, we took a
= 1°. In addition, the isotropy of road surfaces was accepted for d less than 20°.
The indirect measurement of q is carried out for 20 values of b between 0° and 180° and 29 values of tan c between 0 and 12.
E.3 Phase C1: Process analysis

E.3.1 Error cause budget
Taking into account the functional relationships of q and the intermediate quantities Eÿ and Er, the causes of errors taken into account appear on
the diagram in Figure E.4, in which:
This represents the calibration corrections and CS the surface correction;
D represents the distance from the target to the sample.
Figure E.4 — Error cause diagram

FD X 07-021 — 44 —
E.4 Phase C2: Process modeling
Since we are in the case of indirect measurement, the input quantities of the model will be at the same time the intermediate
quantities, the calibration and surface corrections, the repeatability or reproducibility of the measurements,
the influencing parameters.
q = f(, Eÿ S, Er,
d, a, c, Ce, CS, H, D, stability of Eÿ)
The explicit writing of the model is done in two steps.
1st step : physical law of q :

2
Erd
-----------------------------------
q=
SEÿcoscsina
2nd step : modeling of each intermediate quantity when more than one cause of error occurs:
{ Eÿ = Eÿ + Ce + c + stability of Eÿ
{ Er = Ce + H + D
{ Sÿ = +SCe + CS}
E.5 Phase D1: Processing of measurements
E.5.1 Estimation of standard uncertainties

They are given (in relative values), as well as the way in which they are determined, in Table D.1.
Table E.1 — Estimated standard uncertainties for each component

Sizes Standard uncertainties
intermediaries Components Ways to get them %
Eÿ • Luxmeter calibration Calibration certificate uce = 1
• Reproducibility Measurements — extent uEÿ = 0.3

• Angle c Measurements — extent uc = 0.8
• Illumination stability Measurements — experimental ust = 0.1
standard deviation
Er • Luxmeter calibration Calibration certificate uCe = 1

• Sample distance Rectangular law uD = 0.2
to the centering rod
• Horizontality of the sample Rectangular law uH = 0.12
S • Measuring tool Rectangular law uCe = 1.6
• Repeatability Measurements negligible

• Surface correction Measurements — standard deviation ÿ 55° ÿ uCs negligible > 55°
experimental ÿ uCs = 2.3
d • Measuring tool Rectangular law ud = 0.2
a (a = 1°) Rectangular law au = 2
c (c = 85°) Rectangular law negligible uc
1
-----
Uncertainty due to the restriction of the application of the Er law 2 (point source) is assumed to be negligible.
d
The analysis of the causes of errors on each of the intermediate quantities shows that the standard uncertainties are
all uncorrelated.
— 45 — FD X 07-021
E.6 Phase D2: Uncertainty propagation laws

E.6.1 Intermediate quantities
The models of the intermediate quantities Eÿ, ER and S are additive linear; the sensitivity coefficients are
all equal to unity.
2 2 2 2
ÿu = +++
uCe uc 2 ust
uEÿ Eÿ
= ++0.82
0.32 12 + 0.12 = 1.3%
uEÿ
2 2 2 2
ÿ = uCe + +uH uD
uEr
= +
12 0.122 0.22+ = 1%
uEr
2 2 2 2
ÿu us = + +s uCe uCs
For
c ÿ 55°, us = 1.62 + + 0 0 = 1.6%
c > 55°, us = 1.62 2.32 + + 0 = 2.8%
E.6.2 Physical law of q
2
ÿ
Er d
-------------------------------------------
q=
S Eÿ ÿ ÿ cosc sina
2 2 2 2 2 2 2 2 2 2 2 2 2
uc ( ) q = C1 uEÿ +++++
C2 uEr C3 ud C4 us C5 au C6 uc
with :
ÿq
--------- q
=------
C1
= ÿEÿ Eÿ
ÿq
-------- q
= -----
C2
= ÿEr Er
ÿq
----- 2q
=------
C3 =
ÿd d
ÿq
----- = --q
C4 =
ÿS S
ÿq
----- q ÿ cosa
= --------------------
C5 =
ÿa sina
ÿq
----- qÿsinc
= ------------------
C6 =
ÿc cosc
NOTE Negative signs are not taken into account in partial derivatives.
FD X 07-021 — 46 —
The relative composite standard uncertainty on q is thus:
For c ÿ 55°
uc( ) q =
------------- +
12 1.32 + 1.62 22
0.22 + ++ 0 = 3%
q
For c > 55°
uc( ) q =
------------- +
12 1.32 + 2.82 22
0.22 + ++ 0 = 3.8%
q
E.7 Phase D4: Expression of the measurement result and its uncertainty
The measurement result of q is expressed with an uncertainty expanded to k = 2:
U =----------------
--- 2uc( ) q
q q
either
For c ÿ 55° U = 6%
For c > 55° U = 7.6%

— 47 — FD X 07-021
Appendix F
(informative)
Estimation of the uncertainty of the dosage of sodium bromide in a black-
white developer
Init numbering of annex tables [F]!!!
Init numbering of appendix figures [F]!!!
Init numbering of appendix equations [I]!!!
F.1 Phase A1: Definition of the need
The quality of a photographic image depends not only on the intrinsic characteristics of the photographic product on
which it is recorded, but also on the photochemical activity of the developing bath.
(revealer), which, through a redox reaction, makes this image visible (reveal).
Controlling the photochemical activity of developing baths is an important step in the process of controlling the quality
of the products manufactured. The mass concentration of sodium bromide (qNaBr)
is one of the key parameters that is monitored metrologically. For the case presented below, the specification is
3.00 g/l ± 0.25 g/l, with a ratio (half-amplitude of the tolerance interval / Expanded uncertainty) ÿ 4.
F.2 Phase B1: Choice of resources
F.2.1 Principle of measurement

The dosage of sodium bromide is carried out automatically using the classic potentiometry method.
using the equipment shown in Figure F.1 and according to a perfectly defined operating mode, in a room where
the temperature is maintained at 22°C ± 1.5°C.
Br– + Ag+ ÿ AgBr

Legend
1 Control unit 7 Revealer
2 Agitator 8 AgNO3
3 Titroprocessor 4 9 Cruet
Combined Ag/AgCl electrode 10 Volumetric titrator
5 Addition of NaCl 11 Sample changer
6 Silica gel
Figure F.1 — Principle of measurement
The addition of NaCl makes it possible to overcome the variability induced by traces of NaCl released by the products
photographs in the developing baths.
FD X 07-021 — 48 —
F.2.2 Equivalence relationship
CNaBr ÿ VNaBr = CAgNO3 ÿ VAgNO3
%&
qNaBr
---------------
)*
ÿ VNaBr = CAgNO3 ÿ VAgNO3

)*
MNaBr
#$
Or :
CNaBr is the molecular concentration of sodium bromide;
VNaBr is the volume of sodium bromide taken = VRev (revealer);
CAgNO3 is the molecular concentration of silver nitrate;
VAgNO3 is the volume of silver nitrate poured;
qNaBr is the mass concentration of sodium bromide;
MNaBr is the molar mass of sodium bromide.
Either
%& ÿ ÿ
CAgNO3 VAgNO3 MNaBr # $

= ------------------------------------------------- -----------------
qNabr
VRev
F.3 Phase B2: Preliminary analysis of the process: causes of errors

The causes of errors taken into account appear in the diagram in Figure F.2.
Figure F.2 — Error cause diagram

— 49 — FD X 07-021
F.4 Phase D2: Law of propagation of uncertainty
2 2u 2u
2 2%&
%
uc # q $ NaBr
& %&
VS %&V u & u V
# $ AgNO3 # $ AgNO3 # MNaBr %
$ # $ Rev
------------------------ = ----------------------------- + ++
-----------------------------
---------------------- --------------------------
2 2 2 2 2
qNaBr CAgNO3 VAgNO3 MNaBr VRev
F.4.1 Estimation of standard uncertainties
%&
u C# $ AgNO3
The silver nitrate solution is made in the laboratory and its molecular concentration is determined by a
potentiometric dosage with potassium bromide, using the same equipment as previously, at
except for the sample changer.
%&
mKBr PKBr ÿ ÿ 1,000 # $
= ------------------------------------------------- ---- = 0.099 74 mol/l
CAgNO3
VAgNO3 MKBr ÿ
Or :
mKBr is the mass of weighed potassium bromide;
is the volume of silver nitrate poured;

VAgNO3
MKBr is the molar mass of potassium bromide;
PKBr is the purity of potassium bromide.
F.4.2 Combination of relative variances
2 u %&
2%& 2% 2 u % &V
2
u u p & u &
VS#$ AgNO3 # $ mKBr # $ KBr # $ AgNO3 $ MKBr %
#
----------------------------- = ++----------------------- ----------------------------
+ ------------------------ -----------------------------
2 2 2 2 2
CAgNO3 mKBr pKBr VAgNO3 MKBr
The relative variances are detailed in Tables F.1.a), F.1.b), F.1.c) and F.1.d).
FD X 07-021 — 50 —
Table F.1.a) — Estimated standard uncertainties for each component of CAgNO3
Uncertainty component Estimation method Standard uncertainty
Mass KBr weighed — balance characteristics

(130mg)
- display resolution: 0.1 mg (rectangular law) ur 0.1x 2 12 ÿ mg
has)
- drift: negligible udr 0 mg
- temperature sensitivity: 0.003 mg/°C (k = 2) ust (1.5 × 0.003)/2 mg

b)
— experimental reproducibility: negligible uex 0 mg
— standard mass: (accuracy deviation) uj 0.007 5 mg
— air thrust correction difference: negligible udp 0 mg
— vibrations/air movement: negligible uvma 0 mg
22222 2
u(mKBr) ur +++++++
2 udr ust uex uj udp uvma 0.028 9 mg
KBr purity (99.75%) — manufacturer data: purity at least 99.5% (triangular law) up = 0.002 5/ 6
Molar mass KBr — Handbook table

(109.002 3 g/mol)
- K ± 0.1 mg/mol (rectangular law) 3
uK 0.1/ mg/mol
- Br ± 1 mg/mol (rectangular law) 3

uBr 1/ mg/mol
22
u(MKBr) UK + uBr 0.58 mg/mol
Volume AgNO3 — cruet

(10,959 ml)
- “absolute error”: ± 0.03 ml (rectangular law) ub 0.03/ 3 ml
- reproducibility:
b)
— detection of the equivalence point: uex 8.25 × 10-3 ml
— electrode state:
— presence of AgCl in KBr ucl 0.006 ml
— losses of connections and pipes: negligible urt 0 ml
— temperature, volume dilution of water (k = 2) udv (11 × 1.5 × 2.1 × 10-4)/2 ml
22222
ub + +++
% & u V#
uex ucl urt udv 0.020ml
$ AgNO3
2 2 2
%&
u %& u %p & 2u V &
u &
u C#$ AgNO3 # $ mKBr
----------------------- # $ KBr
++----------------------------
+ ------------------------
% # $ AgNO3
---------------------------- # MKBr %
$
--------------------------- 0.001 84
2 2 2 2
CAgNO3 mKBr pKBr VAgNO3 MKBr
has) Postman 2 the zero adjustment and the tailor-made car are carried out;
b) uex = experimental standard deviation.

— 51 — FD X 07-021
Table F.1.b) — Estimated standard uncertainties for each component of VAgNO3
Burette — “absolute error” ± 0.03 ml (rectangular law) uj 0.03/ml 3
— repeatability
Titroprocessor — detection of the equivalent point
Electrode — aging uex 0.022 7 ml

has)
Sample changer — rinsing quality
Automatic NaCl addition — compensation for NaCl release
Ambient temperature — volume expansion of water (k = 2) Udv (2.98 × 1.5 × 2.1 × 10-4)/2 ml
Fittings/Piping — losses negligible urt
%&
For
u V#$ AgNO3 2 2
2 +++units 2 udv urt
0.028 6ml
uj
VAgNO3
= 2,980 5 ml
has) uex = experimental standard deviation.
Table F.1.c) — Estimated standard uncertainties of each component of VRev
Pipette (10 ml) — calibration (weighing) uet 0.089 µl
has)
Pipette/emptied volume — meniscus adjustment uex 0.003 4 ml
— last drop
Temperature — revealing volume expansion (k = 2) udv (10 × 1.5 × 2.1 × 10-4)/2 ml
Air — rate occluded in the sample negligible au

2 2
u(VRev) for 2 uet
+++
units 2 udv au
0.003 75ml
VRev = 9.997 9 ml
has) uex = experimental standard deviation.
Table F.1.d) — Estimated standard uncertainties of each component of MNaBr
N/A — table ± 6 µg/mol (rectangular law) uNa 6/ 3 µg/mol
Br — table ± 1 mg/mol (rectangular law) 3

uBr 1/ µg/mol
%&
For
u M# $ NaBr 0.577 mg/mol
2
MNaBr = 102.893768 g/mol 2 uNa
+ uBr
FD X 07-021 — 52 —
F.4.3 Composite standard uncertainty
2 2 2
& 2 u % &VS 2 u % &V u & u V &
uc % q # $ AgNO3 # $ AgNO3
# $ NaBr
------------------------ = ----------------------------- + ++
----------------------------- $ MNaBr %
#
-------------------------- %# $ Rev
----------------------
2 2 2 2 2
qNaBr CAgNO3 VAgNO3 MNaBr VRev
uc(qNaBr) = 0.030 1 g/l
F.4.4 Expanded uncertainty
U(qNaBr) = 0.060 g/l (k = 2)
F.5 Phase D4: Expression of the result and its uncertainty

qNaBr = (3.059 ± 0.060) g/l (k = 2)
F.6 Phase E1: Decision

The method can be used.
— 53 — FD X 07-021
Appendix G
(informative)
Estimation of the uncertainty of a temperature measurement in an industrial environment

Init numbering of annex tables [G]!!!
Init numbering of appendix figures [G]!!!
Init numbering of appendix equations [J]!!!
This example illustrates the application of the approach for the choice of measuring equipment (Illustration of the
phase B2).
G.1 Phase A1: Definition of the need
Search for a temperature measurement chain in a production unit in the range 20°C to 75°C.
The tolerance interval is [ÿ 0.8 °C + 0.8 °C] and the ,ratio of half tolerance interval to expanded uncertainty
(k = 2) must be ÿ 4.
G.2 Phase B1: Choice of resources
G.2.1 Principle of measurement

Direct measurement using a measurement chain consisting of a sensor + transmitter + converter + display. Transmitter, converter and
display are remote and placed in a room whose temperature is
maintained at (22 ± 2) °C.
G.2.2 Description of the equipment constituting the measurement chain

•
1 double Pt100 ÿ probe, three wires, 1/3 DIN Class B;
•
1 4-20 mA transmitter;
•
1 analog/digital converter;
• 1 display.
G.3 Phase B2: Preliminary analysis of the process: causes of errors

The causes of errors taken into account appear in the diagram in Figure G.1.
(*) “precision” term indicated by the manufacturer and considered to include stability, response time and drift
in time.
Figure G.1 — Error cause diagram
FD X 07-021 — 54 —
G.4 Phase C2: Modeling
h = reading + corrections (not made);

h = f (probe, transmitter, converter, display).
G.4.1 Method of determination
The determination of the standard uncertainty is made from the study of the manufacturer's data, i.e. a type B
determination.
G.4.2 Working hypothesis

All uncertainty components given by manufacturers are considered equivalent to
two standard deviations and the components linked to the indication of the element were calculated for the maximum value at
calculate. The device is supposed to be checked at 22°C.
Probe, transmitter, converter and display being separate elements placed in different locations on the
process, it is assumed that the correlations between the input quantities are negligible.
G.5 Phase D2: Uncertainty propagation law
2
u 2 ( ) h = +++ 2 2 2
u(probe) u(trans) u(conv) u(ra)
G.5.1 Estimation of standard uncertainties

They are given, as well as the way in which they are determined, in Table G.1.
Table G.1 — Estimated standard uncertainties for each component
Uncertainty
Estimation method Standard uncertainty
component
Pt100 probe — stability

has)
- response time u(probe) 0.076 °C
— drift over time
Transmitter - "precision"
0.05% reading = 0.05 × 75/(100 × 2) u(ptr) 0.018 5°C
— ambient temperature at zero input

1.5 × 10-5 % lect/ °C = 2 × 1 × 10-5 × 75/(100 × 2) u(tztr) 75 × 10-7 °C
— ambient temperature on scale

1.5 x 10-5 % lect/ °C = 2 × 1 × 10-5 × 75/(100 × 2) u(tet) 75 × 10-7 °C
— ambient temperature on output

0.01% lect/°C = 2 × 10-4 × 75/(100 × 2) u(tstr) 2.8 × 10-5 °C
2 2 2
u(ptr) 2 +u(tztr)
++ u(tet) u(tstr) u(trans) 0.019 °C
(to be continued)
— 55 — FD X 07-021
Table G.1 — Estimated standard uncertainties of each component (end)
Uncertainty
Estimation method Standard uncertainty
component
Converter — “precision”
0.1% span = 0.1 × 55/(100 × 2) u(pc) 0.027 5°C
— ambient temperature: 50 ppm reading / °C u(c) 0.002 5°C
2
+ 2 u(conv) 0.028 °C
u(pc) u(tc)
Display — display resolution: 0.01 °C (rectangular law) u(ra) 0.003°C
type Experimental
installed another
standard
chain deviation
of on a measurement.
deduced from a probe a) of the same
G.5.2 Composite standard uncertainty
2 2 2 2
uc(h) = +++
u(probe) u(trans) u(conv) u(ra)
uc(h) = 0.0762 0.0192 0.0282 0.0032 +++
uc(h) = 0.085 °C
G.5.3 Expanded uncertainty
U = 0.17°C (k = 2)
G.6 Phase D4: Expression of the result and its uncertainty

h = (reading ± 0.17) °C (k = 2)
G.7 Phase E1: Decision

The tolerance interval IT is [– 0.8 °C, + 0.8 °C], the ratio of the half tolerance interval to the uncertainty
expanded is:
0.8°C
-------------------
= 4.7
0.17°C
It is therefore greater than 4 as was required; the measuring chain is suitable and can be purchased.
FD X 07-021 — 56 —
Appendix H
(informative)
Estimation of the uncertainty of a measurement of the filterability limit temperature

fuels for diesel engines and domestic fuels
Init numbering of annex tables [H]!!!

Init numbering of appendix figures [H]!!!
Init numbering of appendix equations [K]!!!
This example illustrates the case where the process cannot be modeled and where fidelity values are used.
H.1 Phase A1: Definition of the need
The filterability limit temperature is the highest temperature at which a determined volume of fuel ceases to pass through a standardized
filtration device in a limited time, when it is subjected to cooling under standardized conditions.
This measurement is used to assess the lowest temperature to which a fuel will flow freely. In the case of diesel fuels,
the results are generally close to the temperature of
failure in service. Domestic heating systems are, in general, less sensitive and often operate satisfactorily at
temperatures somewhat lower than those determined by the test.
H.2 Phase B1: Choice of resources
The measurement of the Filterability Limit Temperature (TLF) is the subject of standard NF EN 116.
H.3 Phase C3: Interlaboratory tests
An interlaboratory trial was organized in 1996 by a professional monitoring committee. 36 laboratories

using either an automatic device or a manual device (both authorized by standard NF EN 116) have
participated in this trial, which did not result in any elimination of results. The fidelity values obtained during
this test were:
— repeatability limit — r = 1.9°C

reproducibility limit R = 4.4°C
for a laboratory average equal to – 19.1 °C, while the precision values announced by the standard
NF EN 116 at this level are:
— repeatability limit — r = 1.6°C

reproducibility limit R = 4.5°C
— 57 — FD X 07-021
H.4 Phase D3: Use of fidelity values
The interlaboratory test shows that the precision values of the NF EN 116 standard are realistic and representative of a set
of laboratories practicing according to it.
A laboratory having participated in this test, having not been rejected by the elimination tests, therefore having demonstrated
its ability to correctly carry out the test, can claim a standard uncertainty equal to the deviation -type of reproducibility is:
R
= -------
uc( ) y
2.8
Or :
R is the reproducibility of the NF EN 116 standard at the measurement level, since the precision values (known and
published in the standard) vary depending on the measurement level.
Thus for a TLF measurement close to – 19 °C, uc(y) = 1.6 °C and the expanded uncertainty U = 3.2 °C (k = 2).
H.5 Phase E1: Decision — Use of the result and its uncertainty
The previously made estimate of the expanded uncertainty (U = 2R / 2.8 ÿ R 2) must be considered as/ an upper bound of the uncertainty associated
with an individual and one-off laboratory measurement, but it can be considered as representative of the uncertainty associated with measurements
carried out by this laboratory, in a context of “routine” control where, over time, various operators will carry out measurements on products whose nature
and TLF level vary.
FD X 07-021 — 58 —
Bibliography
[1] NF EN 116, Fuels diesel engines domestic heating installations — Determining the filterability limit temperature
(classification index: M 07-042).
[2] NF EN 22719, And

Lubricating petroleum products — Determination of flash point — Pensky-Martens method
enclosed in a(classification
vase index: M 07-019).
[3] ISO 4012, [4] Concrete — Determination of compressive strength of specimens .
The of uncertainty concept ace applied chemical measurements — W Horwitz and R. Albert Analystto ,
June 1997 — vol 122 (615-617)

FD X07 021 Métrologie Et Application de La Statistique

Uploaded by

Copyright:

Available Formats

FD X07 021 Métrologie Et Application de La Statistique

Uploaded by

Document Information

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

FD X07 021 Métrologie Et Application de La Statistique

Uploaded by

Copyright:

Available Formats

Machine Translated by Google

Classification index: X 07-021

ICS: 03.120.30; 17.020

E: Fundamental standards — Metrology and application of the statistics —

or international documents dealing with the same subject.

Descriptors International Technical Thesaurus: statistics, metrology, measurement, testing,

© AFNOR 1999 AFNOR 1999 1st drawing 99-10-F

Metrology in the company AFNOR X07B

Members of the standardization commission

M ALLIOUZ ESSO SAF

Statistical methods AFNOR X06E

Members of the standardization commission

Members of the “Measurement uncertainty” working group

M BELLON CENTRAL LABORATORY OF ELECTRICAL INDUSTRIES (LCIE)

Foreword .............................................. .................................................. .................................................. ... 5

1 Application domain .............................................. .................................................. ............................. 5

2 Normative references .................................................. .................................................. .................. 6

3 Terms and definitions .............................................. .................................................. ....................... 6

5 Comments on the developed flowchart ......................................... ................................... 12

6 Appendices .................................................. .................................................. ......................................... 23

Annex A (informative) Estimation of the calibration uncertainty of a liquid expansion

Annex B (informative) Estimation of the uncertainty of a measurement of polluting CO emissions

Annex C (informative) Estimation of uncertainty associated with height measurement

Annex D (informative) Estimation of the uncertainty of a current intensity measurement .................................. 34

Annex E (informative) Estimation of the uncertainty of a luminance coefficient measurement

Annex F (informative) Estimation of uncertainty of sodium bromide determination

Annex G (informative) Estimation of the uncertainty of a temperature measurement in an industrial environment 53

Bibliography .................................................. .................................................. .................................................. 58

General Commission “Metrology in the Commission of

Standardization “Statistical methods” (X06E)).

This provides the scientific metrological bases 4 text and for

of the reader therefore reproduced in this document.

difference between the results from a test of a nature or measurement.

A set of operations aimed atmeasurement

A test the technical

Or service given, depending on A

result. or uncertainties in test results. to estimates

method, analysis, in order

the and the one which completely technical

NF EN 30012-1, Quality assurance requirements for measuring equipment — Part 1: Confirmation

NF EN ISO 14253-1, Geometric specification (GPS) — Verification of product parts and

(classification index: E 10-201-1) .

NF ISO 3534–1, and 1: and

NF ISO 10012-2, Insurance Measurement

(classification index: X 30-204)

NF X 07-001, Core Standards — International Vocabulary general core terms of and

NF X 07-010, Metrology — metrological function in the company.

3 Terms and definitions

NOTE 1 These conditions are called repeatability conditions.

NOTE 2 Repeatability conditions include:

— same operating procedure;

NOTE 2 The conditions that are varied may include:

[NF ISO 3534-1, § 3.14]

NOTE See 3.14, NOTE 3.

[NF ISO 3534-1, § 3.16]

[NF ISO 3534-1, § 3.20]

[NF ISO 3534-1, § 3.21]

4 Approach to the estimation, expression and use of uncertainty

— Expression of the need for measurement or testing Phase A;

— Measuring or testing process Phase C