Determination of the characteristic limits (decision threshold, detection limit and limits of the confidence interval) for measurements of ionizing radiation - Fundamentals and application

ISO 11929:2010 specifies a procedure, in the field of ionizing radiation metrology, for the calculation of the "decision threshold", the "detection limit" and the "limits of the confidence interval" for a non-negative ionizing radiation measurand, when counting measurements with preselection of time or counts are carried out, and the measurand results from a gross count rate and a background count rate as well as from further quantities on the basis of a model of the evaluation. In particular, the measurand can be the net count rate as the difference of the gross count rate and the background count rate, or the net activity of a sample. It can also be influenced by calibration of the measuring system, by sample treatment and by other factors.

Détermination des limites caractéristiques (seuil de décision, limite de détection et extrémités de l'intervalle de confiance) pour mesurages de rayonnements ionisants — Principes fondamentaux et applications

L'ISO 11929:2010 spécifie une procédure, dans le domaine de la métrologie des rayonnements ionisants, permettant de calculer le «seuil de décision», la «limite de détection» et les «extrémités de l'intervalle de confiance» d'un mesurande non négatif caractéristique d'un rayonnement ionisant, lorsque des mesurages par comptage avec présélection de temps ou du nombre d'impulsions sont réalisés, et que le mesurande résulte d'un taux de comptage brut et d'un taux de comptage du bruit de fond ainsi que de grandeurs supplémentaires sur la base d'un modèle d'évaluation. En particulier, le mesurande peut être le taux de comptage net défini comme étant la différence du taux de comptage brut et du taux de comptage du bruit de fond, ou l'activité nette d'un échantillon. Il peut également être influencé par l'étalonnage du système de mesure, par le traitement de l'échantillon et par d'autres facteurs.

General Information

Status
Withdrawn
Publication Date
23-Feb-2010
Withdrawal Date
23-Feb-2010
Current Stage
9599 - Withdrawal of International Standard
Start Date
13-Feb-2019
Completion Date
13-Dec-2025
Ref Project

Relations

Standard
ISO 11929:2010 - Determination of the characteristic limits (decision threshold, detection limit and limits of the confidence interval) for measurements of ionizing radiation -- Fundamentals and application
English language
60 pages
sale 15% off
Preview
sale 15% off
Preview
Standard
ISO 11929:2010 - Détermination des limites caractéristiques (seuil de décision, limite de détection et extrémités de l'intervalle de confiance) pour mesurages de rayonnements ionisants -- Principes fondamentaux et applications
French language
61 pages
sale 15% off
Preview
sale 15% off
Preview

Frequently Asked Questions

ISO 11929:2010 is a standard published by the International Organization for Standardization (ISO). Its full title is "Determination of the characteristic limits (decision threshold, detection limit and limits of the confidence interval) for measurements of ionizing radiation - Fundamentals and application". This standard covers: ISO 11929:2010 specifies a procedure, in the field of ionizing radiation metrology, for the calculation of the "decision threshold", the "detection limit" and the "limits of the confidence interval" for a non-negative ionizing radiation measurand, when counting measurements with preselection of time or counts are carried out, and the measurand results from a gross count rate and a background count rate as well as from further quantities on the basis of a model of the evaluation. In particular, the measurand can be the net count rate as the difference of the gross count rate and the background count rate, or the net activity of a sample. It can also be influenced by calibration of the measuring system, by sample treatment and by other factors.

ISO 11929:2010 specifies a procedure, in the field of ionizing radiation metrology, for the calculation of the "decision threshold", the "detection limit" and the "limits of the confidence interval" for a non-negative ionizing radiation measurand, when counting measurements with preselection of time or counts are carried out, and the measurand results from a gross count rate and a background count rate as well as from further quantities on the basis of a model of the evaluation. In particular, the measurand can be the net count rate as the difference of the gross count rate and the background count rate, or the net activity of a sample. It can also be influenced by calibration of the measuring system, by sample treatment and by other factors.

ISO 11929:2010 is classified under the following ICS (International Classification for Standards) categories: 17.240 - Radiation measurements. The ICS classification helps identify the subject area and facilitates finding related standards.

ISO 11929:2010 has the following relationships with other standards: It is inter standard links to ISO 11929-4:2020, ISO 11929-3:2019, ISO 11929-1:2019, ISO 11929-2:2019, ISO 11929-7:2005, ISO 11929-1:2000, ISO 11929-6:2005, ISO 11929-8:2005, ISO 11929-2:2000, ISO 11929-4:2001, ISO 11929-5:2005, ISO 11929-3:2000. Understanding these relationships helps ensure you are using the most current and applicable version of the standard.

ISO 11929:2010 is associated with the following European legislation: EU Directives/Regulations: 2018-01-1150. When a standard is cited in the Official Journal of the European Union, products manufactured in conformity with it benefit from a presumption of conformity with the essential requirements of the corresponding EU directive or regulation.

You can purchase ISO 11929:2010 directly from iTeh Standards. The document is available in PDF format and is delivered instantly after payment. Add the standard to your cart and complete the secure checkout process. iTeh Standards is an authorized distributor of ISO standards.

Standards Content (Sample)


INTERNATIONAL ISO
STANDARD 11929
First edition
2010-03-01
Determination of the characteristic limits
(decision threshold, detection limit and
limits of the confidence interval) for
measurements of ionizing radiation —
Fundamentals and application
Détermination des limites caractéristiques (seuil de décision, limite de
détection et extrémités de l'intervalle de confiance) pour mesurages de
rayonnements ionisants — Principes fondamentaux et applications

Reference number
©
ISO 2010
PDF disclaimer
This PDF file may contain embedded typefaces. In accordance with Adobe's licensing policy, this file may be printed or viewed but
shall not be edited unless the typefaces which are embedded are licensed to and installed on the computer performing the editing. In
downloading this file, parties accept therein the responsibility of not infringing Adobe's licensing policy. The ISO Central Secretariat
accepts no liability in this area.
Adobe is a trademark of Adobe Systems Incorporated.
Details of the software products used to create this PDF file can be found in the General Info relative to the file; the PDF-creation
parameters were optimized for printing. Every care has been taken to ensure that the file is suitable for use by ISO member bodies. In
the unlikely event that a problem relating to it is found, please inform the Central Secretariat at the address given below.

©  ISO 2010
All rights reserved. Unless otherwise specified, no part of this publication may be reproduced or utilized in any form or by any means,
electronic or mechanical, including photocopying and microfilm, without permission in writing from either ISO at the address below or
ISO's member body in the country of the requester.
ISO copyright office
Case postale 56 • CH-1211 Geneva 20
Tel. + 41 22 749 01 11
Fax + 41 22 749 09 47
E-mail copyright@iso.org
Web www.iso.org
Published in Switzerland
ii © ISO 2010 – All rights reserved

Contents Page
Foreword .iv
Introduction.v
1 Scope.1
2 Normative references.1
3 Terms and definitions .2
4 Quantities and symbols.4
5 Fundamentals .5
5.1 General aspects concerning the measurand.5
5.2 Model .6
5.3 Calculation of the standard uncertainty as a function of the measurand .8
6 Characteristic limits and assessments .10
6.1 Specifications .10
6.2 Decision threshold .10
6.3 Detection limit.10
6.4 Limits of the confidence interval .12
6.5 Assessment of a measurement result.13
6.6 Assessment of a measurement procedure.14
7 Documentation .15
Annex A (informative) Overview of the general procedure .16
Annex B (normative) Various applications .18
Annex C (normative) Applications to counting spectrometric measurements.25
Annex D (informative) Application examples.37
Annex E (informative) Distribution function of the standardized normal distribution .49
Annex F (informative) Explanatory notes .51
Bibliography.59

Foreword
ISO (the International Organization for Standardization) is a worldwide federation of national standards bodies
(ISO member bodies). The work of preparing International Standards is normally carried out through ISO
technical committees. Each member body interested in a subject for which a technical committee has been
established has the right to be represented on that committee. International organizations, governmental and
non-governmental, in liaison with ISO, also take part in the work. ISO collaborates closely with the
International Electrotechnical Commission (IEC) on all matters of electrotechnical standardization.
International Standards are drafted in accordance with the rules given in the ISO/IEC Directives, Part 2.
The main task of technical committees is to prepare International Standards. Draft International Standards
adopted by the technical committees are circulated to the member bodies for voting. Publication as an
International Standard requires approval by at least 75 % of the member bodies casting a vote.
Attention is drawn to the possibility that some of the elements of this document may be the subject of patent
rights. ISO shall not be held responsible for identifying any or all such patent rights.
ISO 11929 was prepared by Technical Committee ISO/TC 85, Nuclear energy, Subcommittee SC 2, Radiation
protection.
This first edition of ISO 11929 cancels and replaces ISO 11929-1:2000, ISO 11929-2:2000, ISO 11929-3:2000,
ISO 11929-4:2001, ISO 11929-5:2005, ISO 11929-6:2005, ISO 11929-7:2005 and ISO 11929-8:2005, which
have been technically revised, specifically with reference to the type of statistical treatment of the data.
iv © ISO 2010 – All rights reserved

Introduction
The limits to be provided according to this International Standard by means of statistical tests and specified
probabilities allow detection possibilities to be assessed for a measurand and for the physical effect quantified
by this measurand as follows:
⎯ the “decision threshold” gives a decision on whether or not the physical effect quantified by the
measurand is present;
⎯ the “detection limit” indicates the smallest true value of the measurand which can still be detected with the
applied measurement procedure; this gives a decision on whether or not the measurement procedure
satisfies the requirements and is therefore suitable for the intended measurement purpose;
⎯ the “limits of the confidence interval” enclose, in the case of the physical effect recognized as present, a
confidence interval containing the true value of the measurand with a specified probability.
Hereinafter, the limits mentioned are jointly called “characteristic limits”.
Since measurement uncertainty plays an important part in this International Standard, the evaluation of
measurements and the treatment of measurement uncertainties are carried out by means of the general
procedures according to ISO/IEC Guide 98-3; see also References [1, 2]. This enables the strict separation of
the evaluation of the measurements, on the one hand (Clause 5), and the provision and calculation of the
characteristic limits, on the other hand (Clause 6). This International Standard is based on Bayesian statistics
according to References [6 to 19], such that uncertain quantities and influences, which do not behave
randomly in measurements repeated several times or in counting measurements, can also be taken into
account.
Equations are provided for the calculation of the characteristic limits of an ionizing radiation measurand via the
“standard measurement uncertainty” of the measurand (hereinafter “standard uncertainty”). The standard
uncertainties of the measurement, as well as those of sample treatment, calibration of the measuring system
and other influences are taken into account. However, the latter standard uncertainties are assumed to be
known from previous investigations.

INTERNATIONAL STANDARD ISO 11929:2010(E)

Determination of the characteristic limits (decision threshold,
detection limit and limits of the confidence interval) for
measurements of ionizing radiation — Fundamentals and
application
1 Scope
This International Standard specifies a procedure, in the field of ionizing radiation metrology, for the
calculation of the “decision threshold”, the “detection limit” and the “limits of the confidence interval” for a non-
negative ionizing radiation measurand, when counting measurements with preselection of time or counts are
carried out, and the measurand results from a gross count rate and a background count rate as well as from
further quantities on the basis of a model of the evaluation. In particular, the measurand can be the net count
rate as the difference of the gross count rate and the background count rate, or the net activity of a sample. It
can also be influenced by calibration of the measuring system, by sample treatment and by other factors.
This International Standard also applies, in the same way to:
⎯ counting measurements on moving objects (see B.2);
⎯ measurements with linear-scale analogue count rate measuring instruments (hereinafter called
ratemeters, see B.3);
⎯ repeated counting measurements with random influences (see B.4);
⎯ counting measurements on filters during accumulation of radioactive material (see B.5);
⎯ counting spectrometric multi-channel measurements, if particular lines in the spectrum are to be
considered and no adjustment calculations, for instance, an unfolding, have to be carried out (see C.2 to
C.4);
⎯ counting spectrometric multi-channel measurements if evaluated by unfolding methods (see C.5), in
particular, alpha- and gamma-spectrometric measurements (see C.5.5 and C.5.6, respectively).
This International Standard also applies analogously to other measurements of any kind if the same model of
[18]
the evaluation is involved. In this sense, it is also applicable to measurements with albedo dosimeters .
[21]
Further practical examples can be found in other International Standards, for example ISO 18589 ,
[22] [23] [24] [25] [26] [27] [28]
ISO 9696 , ISO 9697 , ISO 9698 , ISO 9699 , ISO 10703 , ISO 7503 and ISO 28218 .
2 Normative references
The following referenced documents are indispensable for the application of this document. For dated
references, only the edition cited applies. For undated references, the latest edition of the referenced
document (including any amendments) applies.
ISO 31-0, Quantities and units — Part 0: General principles
ISO 31-9, Quantities and units — Part 9: Atomic and nuclear physics
ISO/IEC Guide 98-3:2008, Uncertainty of measurement — Part 3: Guide to the expression of uncertainty in
measurement (GUM:1995)
ISO/IEC Guide 99:2007, International vocabulary of metrology — Basic and general concepts and associated
terms (VIM)
ISO 3534-1, Statistics — Vocabulary and symbols — Part 1: General statistical terms and terms used in
probability
3 Terms and definitions
For the purposes of this document, the terms and definitions given in ISO 31-0, ISO 31-9, ISO/IEC Guide 98-3,
ISO/IEC Guide 99 and ISO 3534-1 and the following apply.
3.1
measurement procedure
set of operations, described specifically, used in the performance of particular measurements according to a
given method
[ISO/IEC Guide 99:2007, 2.6]
3.2
measurand
particular quantity subject to measurement
[ISO/IEC Guide 99:2007, 2.3]
NOTE In this International Standard, a measurand is non-negative and quantifies a nuclear radiation effect. The
effect is not present if the true value of the measurand is zero. An example of a measurand is the intensity of an energy
line in a spectrum above the background in a spectrometric measurement.
3.3
result of a measurement
value attributed to a measurand, obtained by measurement
[ISO/IEC Guide 99:2007, 2.9]
3.4
uncertainty of measurement
uncertainty
non-negative parameter, which characterizes the dispersion of the values which could reasonably be
attributed to the measurand
[ISO/IEC Guide 99:2007, 2.26]
See also ISO/IEC Guide 98-3.
NOTE The uncertainty of a measurement derived according to ISO/IEC Guide 98-3 comprises, in general, many
components. Some of these components can be evaluated from the statistical distribution of the results of series of
measurements and can be characterized by experimental standard deviations. The other components, which can also be
characterized by standard deviations, are evaluated from assumed or known probability distributions based on experience
and other information.
3.5
model of evaluation
set of mathematical relationships between all measured and other quantities involved in the evaluation of
measurements
NOTE The model of evaluation does not need to be an explicit function; it can also be an algorithm realized by a
computer code.
2 © ISO 2010 – All rights reserved

3.6
decision threshold
value of the estimator of the measurand, which when exceeded by the result of an actual measurement using
a given measurement procedure of a measurand quantifying a physical effect, one decides that the physical
effect is present
NOTE 1 The decision threshold is defined such that in cases where the measurement result, y, exceeds the decision
threshold, y*, the probability that the true value of the measurand is zero is less or equal to a chosen probability, α.
NOTE 2 If the result, y, is below the decision threshold, y*, the result cannot be attributed to the physical effect;
nevertheless it cannot be concluded that it is absent.
3.7
detection limit
smallest true value of the measurand which ensures a specified probability of being detectable by the
measurement procedure
NOTE With the decision threshold according to 3.6, the detection limit is the smallest true value of the measurand for
which the probability of wrongly deciding that the true value of the measurand is zero is equal to a specified value, β, when,
in fact, the true value of the measurand is not zero.
3.8
limits of the confidence interval
values which define a confidence interval containing the true value of the measurand with a specified
probability
NOTE A confidence interval is sometimes known as a credible interval or a Bayesian interval. It is characterized in
this International Standard by a specified probability (1− γ ) .
3.9
best estimate of the true value of the measurand
expectation value of the probability distribution of the true value of the measurand, given the experimental
result and all prior information on the measurand
NOTE The best estimate, among all possible estimates of the measurand on the basis of given information, which is
associated with the minimum uncertainty.
3.10
guideline value
value which corresponds to scientific, legal or other requirements and which is intended to be assessed by the
measurement procedure
NOTE 1 The guideline value can be given, for example, as an activity, a specific activity or an activity concentration, a
surface activity or a dose rate.
NOTE 2 The comparison of the detection limit with a guideline value allows a decision on whether or not the
measurement procedure satisfies the requirements set forth by the guideline value and is therefore suitable for the
intended measurement purpose. The measurement procedure satisfies the requirement if the detection limit is smaller
than the guideline value.
3.11
background effect
measurement effect caused by radiation other than that caused by the object of the measurement itself
EXAMPLE Natural radiation sources.
3.12
net effect
contribution of the possible radiation of a measurement object (for instance, of a radiation source or radiation
field) to the measurement effect
3.13
gross effect
measurement effect caused by the background effect and the net effect
3.14
shielding factor
factor describing the reduction of the background count rate by the effect of shielding caused by the
measurement object
3.15
relaxation time constant
duration in which the output signal of a linear-scale ratemeter decreases to 1/e times the starting value after
stopping the sequence of the input pulses
3.16
background
〈spectrometric measurements〉 number of events of no interest in the region of a specific line in the spectrum
NOTE The events can be due to the background effect by the environmental radiation and also to the sample itself
(for instance, from other lines).
4 Quantities and symbols
The symbols for auxiliary quantities and the symbols only used in the annexes are not listed. Physical
quantities are denoted by upper-case letters but shall be carefully distinguished from their values, denoted by
the corresponding lower-case letters. In addition, the special quantities and symbols for unfolding in
spectrometric measurements given in C.5.1 and for Bayesian statistics given in F.2.1 are used.
m number of input quantities
X input quantity (1im=,., )
i
x estimate of the input quantity X
i i
ux() standard uncertainty of the input quantity X associated with the estimate x
i i i
hx() standard uncertainty ux() as a function of the estimate x
11 1 1
∆x width of the region of the possible values of the input quantity X
i i
uw() relative standard uncertainty of a quantity Y associated with the estimate w
rel k
G model function
Y random variable as an estimator of the measurand; also used as the symbol for the non-negative
measurand itself, which quantifies the physical effect of interest
 
y true value of the measurand; if the physical effect of interest is not present, then y = 0; otherwise,
y > 0
y determined value of the estimator Y, estimate of the measurand, primary measurement result of the
measurand
y values y from different measurements ( j = 0,1, 2, .)
j
uy() standard uncertainty of the measurand associated with the primary measurement result y
4 © ISO 2010 – All rights reserved

 
uy() standard uncertainty of the estimator Y as a function of the true value y of the measurand
ˆ
y best estimate of the measurand
uy()ˆ standard uncertainty of the measurand associated with the best estimate yˆ
y* decision threshold of the measurand
#
y detection limit of the measurand
#

y approximations of the detection limit y
i
y guideline value of the measurand
r
 
y , y lower and upper limit of the confidence interval, respectively, of the measurand
ρ count rate as an input quantity X
i i
ρ count rate of the net effect (net count rate)
n
ρ count rate of the gross effect (gross count rate)
g
ρ count rate of the background effect (background count rate)
n number of counted pulses obtained from the measurement of the count rate ρ
i i
n , n number of counted pulses of the gross effect and of the background effect, respectively
g 0
t measurement duration of the measurement of the count rate ρ
i i
t , t measurement duration of the measurement of the gross effect and of the background effect,
g 0
respectively
r estimate of the count rate ρ
i i
r , r estimate of the gross count rate and of the background count rate, respectively
g 0
τ , τ relaxation time constant of a ratemeter used for the measurement of the gross effect and of the
g 0
background effect, respectively
α, β probability of the error of the first and second kind, respectively
1− γ probability for the confidence interval of the measurand
k , k quantiles of the standardized normal distribution for the probabilities p and q, respectively (for
p q
instance 1p=− α , 1− β or 12− γ )
Φ()t distribution function of the standardized normal distribution; Φ=()kp applies
p
5 Fundamentals
5.1 General aspects concerning the measurand
A non-negative measurand shall be assigned to the physical effect to be investigated according to a given

measurement task. The measurand shall quantify the effect. It assumes the true value y = 0 if the effect is not
present in a particular case.
Then, a random variable Y, an estimator, shall be assigned to the measurand. The symbol Y is also used in
this clause for the measurand itself. A value y of the estimator Y, determined from measurements, is an
estimate of the measurand. It shall be calculated as the primary measurement result together with the primary
standard uncertainty uy( ) , of the measurand associated with y. These two values form the primary complete
measurement result for the measurand and are obtained in accordance with ISO/IEC Guide 98-3 (see also
References [1, 2]) by evaluation of the measurement data and other information by means of a model (of the
evaluation), which mathematically connects all the quantities involved (see 5.2). In general, the fact that the
measurand is non-negative is not explicitly taken into account in the evaluation. Therefore, y may be negative,

especially when the measurand nearly assumes the true value y = 0. The primary measurement result, y,
ˆ ˆ
differs from the best estimate, y , of the measurand calculated in 6.5. With y , the knowledge that the
measurand is non-negative is taken into account. The standard uncertainty, uy()ˆ , associated with yˆ is
smaller than uy( ) .
5.2 Model
5.2.1 General model
In many cases, the measurand, Y, is a function of several input quantities, X , in the form of Equation (1):
i
YG= (X , .,X ) (1)
1 m
Equation (1) is the model of the evaluation. Substituting given estimates, x , of the input quantities, X , in the
i i
model function, G, Equation (1) yields the primary measurement result y of the measurand as:
yG= (x , .,x ) (2)
1 m
The standard uncertainty, uy( ) , of the measurand associated with the primary measurement result, y, is
calculated according to Equation (3), if the input quantities, X , are independently measured and standard
i
uncertainties, (ux) , associated with the estimates, x , are given, from the relation:
i i
m
⎛⎞
∂G
uy() = u (x) (3)
⎜⎟
∑ i
∂X
⎝⎠i
i=1
In Equation (3), the estimates, x , shall be substituted for the input quantities, X , in the partial derivatives of
i i
G. The determination of the estimates, x , and the associated standard uncertainties, ux( ), and also the
i i
numerical or experimental determination of the partial derivatives are in accordance with ISO/IEC Guide 98-3
or References [1, 2]. For a count rate, X = ρ , with the given counting result, n , recorded during the
ii i
measurement of duration, t , the specifications xr==nt and ux()==n t r t apply (see also F.1).
i ii i i iii ii
If the input quantities are not independently measured and for more complicated measurement evaluations
such as unfolding, see C.5.2.
In 5.2.2, the input quantity, X , for instance the gross count rate, is taken as that quantity whose value, x , is
1 1

not given when a true value, y , of the measurand, Y, is specified within the framework of the calculation of the
decision threshold and the detection limit. Analogously, the input quantity, X , is assigned in a suitable way
to the background effect. The data of the other input quantities are taken as given from independent previous
investigations.
5.2.2 Model in ionizing radiation measurements

In this International Standard, the measurand, Y, with its true value, y , relates to a sample of radioactive
material and is determined from counting the gross effect and the background effect with preselection of time
or counts. In particular, Y can be the net count rate, ρ , or the net activity, A, of the sample. The symbols
n
belonging to the counting of the gross effect and of the background effect are marked in the following by the
subscripts g and 0, respectively.
6 © ISO 2010 – All rights reserved

In this International Standard, the model is specified as follows:
XX ⋅⋅⋅
Y==G(X , .,X ) (XX− X−X )⋅ = (XX− X−X )⋅W (4)
11m 234 1234
XX ⋅⋅⋅
with
XX ⋅⋅⋅
W = (5)
XX ⋅⋅⋅
X = ρ is the gross count rate and X = ρ is the background count rate. The other input quantities, X ,
1g 20 i
are calibration, correction or influence quantities, or conversion factors, for instance the emission or response
probability or, in particular, X is a shielding factor and X an additional background correction quantity. If
3 4
some of the input quantities are not involved, x = 1 (3ii=>; 4) , x = 0 and ux( ) = 0 shall be set for them.
i 4 i
For the count rates, xr==n t and ux()==n t r t as well as xr==n t and
1g g g 1gg gg 20 0 0
ux()==n t r t apply.
200 00
By substituting the estimates, x , in Equation (4), the primary estimate, y, of the measurand, Y, gives the
i
result:
⎛⎞n
n
g
y==G(x , .,x ) (x−xxx− )⋅w= (r−rx−x )⋅w=⎜⎟− xx−⋅w (6)
1 m 1 23 4 g 03 4 3 4
⎜⎟
tt
g0
⎝⎠
with
xx ⋅⋅⋅
w = (7)
xx ⋅⋅⋅
With the partial derivatives:
∂G ∂G ∂G ∂G ∂GY
= W ; =−XW ; =−XW ; =−W ; =± (5iW) (8)
3 2
∂X ∂X ∂X ∂X ∂XX
1 2 3 4 ii
and by substituting the estimates x , w and y, Equation (3) yields the standard uncertainty uy( ) of the
i
measurand associated with y:
2 2 22 22 2 2 2
⎡⎤
uy()= w⋅ u (x )+++x u (x ) x u (x ) u (x )+y u (w)
13 2 2 3 4 rel
⎣⎦
(9)
22 22 2 22
⎡⎤
=⋅wrt+xrt+ru()x+u(x)+yu (w)
gg 30 0 0 3 4 rel
⎣⎦
where
m 2
ux()
2 i
uw() = (10)
rel ∑
x
i=5 i
is the sum of the squared relative standard uncertainties of the quantities X to X . For m < 5, the values
5 m
w = 1 and uw() =0 apply.
rel
The estimate x and the standard uncertainty ux( ) of X (3im= ,., ) are taken as known from previous
i i i
investigations or as values of experience according to other information. In the previous investigations, x can
i
be determined as an arithmetic mean value and ux() as an empirical variance (see B.4.1). If necessary,
i
ux() can also be calculated as the variance of a rectangular distribution over the region of the possible
i
values of X with the width ∆x . This yields ux()=∆(x) 12.
i i ii
For the application of the procedure to particular measurements, including spectrometric measurements, see
Annexes B and C.
5.3 Calculation of the standard uncertainty as a function of the measurand
5.3.1 General aspects
For the provision and numerical calculation of the decision threshold in 6.2 and of the detection limit in 6.3, the
standard uncertainty of the measurand is needed as a function uy( ) of the true value yW 0 of the measurand.
This function shall be determined in a way similar to uy( ) within the framework of the evaluation of the
measurements by application of ISO/IEC Guide 98-3; see also References [1, 2]. In most cases, uy( ) shall be

formed as a positive square root of a variance function uy() calculated first. This function shall be defined,
unique and continuous for all yW 0 and shall not assume negative values.

In most cases, uy( ) can be explicitly specified, provided that ux() is given as a function hx() of x . In such
1 11 1
cases, y shall be formally replaced by y and Equation (2) shall be solved for x . With a specified y , the
value x can also be calculated numerically from Equation (2); for instance, by means of an iteration
procedure, which results in x as a function of y and x , ., x . This function shall replace x in
1 2 m 1

Equation (3) and in ux() =h(x), which finally yields uy( ) instead of uy( ) . In the case of the model
according to Equation (6) and 5.3.2, one shall proceed in this way. Otherwise, 5.3.3 shall be applied, where

uy() follows as an approximation by interpolation from the data y and uy( ) of several measurements.
j j
5.3.2 Explicit calculation
When, in the case of the model according to Equation (6), the standard uncertainty, ux() , of the gross count
rate X = ρ , is given as a function hx() of the estimate, xr= , either hx() = x t or
1g 11 1g 11 1 g
hx() =x n applies if the measurement duration, t (time preselection), or, respectively, the number, n ,
11 1 g g g
of recorded pulses (preselection of counts) is specified.

The value y shall be formally replaced by y . This allows the elimination of x in the general case and, in
particular, of n with time preselection and of t with preselection of counts in Equation (9) by means of
g g

Equation (6). These values are not available when y is specified. This yields in the general case according to
Equation (6)
xy=+w xx+x (11)
By substituting x according to Equation (11) in the given function hx() , i.e. with
1 11
ux()=+h(yw xx+x ) , the following results from Equation (9):
11 23 4
2 2 22 22 2 2 2
⎡⎤
uy()=⋅w h(y w+x x+x)+x u()x+x u()x+u(x)+y u (w) (12)
12343 22 3 4 rel
⎣⎦
With time preselection and because of xn= t and xr= ,
1g g 20
nt=⋅()yw+rx +x (13)
gg 03 4
8 © ISO 2010 – All rights reserved

is obtained from Equation (11). Then, with hx()==x t n t and by substituting n according to
11 1 g g g g
Equation (13) and with ux() =r t , Equation (12) leads to
⎡⎤
  
uy()=⋅w (y w+r x+x ) t+x r t+r u (x )+u (x )+y u (w) (14)
03 4 g 3 0 0 0 3 4 rel
⎣⎦
With preselection of counts,
n
g
t = (15)
g

yw++r x x
03 4
is analogously obtained. Then, with hx()==x t n t and by substituting t according to Equation (15)
11 1 g g g g
and with ux() =r n , Equation (12) leads to
20 0
⎡⎤
uy()=⋅w (y/w+r x+x ) n+x r n+r u (x )+u (x )+y u (w) (16)
03 4 g 3 0 0 0 3 4 rel
⎣⎦
#
Equation (22) has a solution, which is the detection limit, y , if, with time preselection, the following condition
is satisfied:
ku ()w <1 (17)
1r−β el
or with preselection of counts, the following condition is satisfied:
ku⋅+ ()w<1 (18)
1−β
rel
n
g
Otherwise, it can happen that a detection limit does not exist because of too great an uncertainty of the
quantities X to X , summarily expressed by uw(). The condition according to Equation (17) also applies
5 m rel
in the case of Equation (12) if hx() increases for growing x more slowly than x , i.e. if hx()x →0 for
11 1 1 11 1
x →∞ .
5.3.3 Approximations

It is often sufficient to use the following approximations for the function uy( ) , in particular, if the standard
uncertainty, ux(), is not known as a function hx(). A prerequisite is that measurement result, y , and
1 11 j
associated standard uncertainties, uy( ) , calculated according to 5.1 and 5.2 from previous measurements of
j
the same kind, are already available ( j = 0,1,2,.). The measurements shall be carried out on different
samples with differing activities, but in other respects as far as possible under similar conditions. One of the

measurements can be a background effect measurement or a blank measurement with y =0 and, for

instance, 0j = . Then, y = 0 shall be set and uu(0) = (y ) . The measurement currently carried out can be
0 0
taken as a further measurement with j = 1.
 
The function uy( ) often shows a rather slow increase. Therefore, the approximation uy() =uy( ) is sufficient
in some of these cases, especially if the primary measurement result, y , of the measurand is not much
larger than the associated standard uncertainty uy() .
If only uu(0) = (y ) , y > 0 and uy() are known, the following linear interpolation often suffices:
0 1 1
22 2
uy()=−u (0)(1 yy )+u (y )yy (19)
11 1
If the results y , y and y as well as the associated standard uncertainties uy() , uy() , and uy() from
0 1 2 0 1 2
three measurements are available, the following bilinear interpolation can be used:
  
()yy−−(yy) ()yy−−()yy (y−y)(y−y)
22 2 2
12 02 0 1

uy()=⋅uy() +uy( )⋅ +uy()⋅ (20)
01 2
(yy−−)(yy) ()y−−y()y y (yy−)(yy−)
01 0 2 1 0 1 2 2 0 2 1

If results from many similar measurements are given, the parabolic shape of the function uy() can also be
determined by an adjustment calculation.
6 Characteristic limits and assessments
6.1 Specifications
The probability, α, of the error of the first kind, the probability, β, of the error of the second kind and the
probability, 1− γ , for the confidence interval shall be specified. The choice depends on the application. A
frequently cited choice is α = β and the value 0,05 for α and β. Then, kk== 1, 65 . If the value of 0,05
11−−αβ
is chosen for γ, then k = 1, 96 (see Annex E).
12−γ
If it is to be assessed whether or not a measurement procedure for the measurand satisfies the requirements
to be fulfilled for scientific, legal or other reasons (see 6.6), a guideline value, y , as a value of the measurand,
r
for instance, an activity, shall also be specified.
6.2 Decision threshold
The decision threshold, y* , of the non-negative measurand according to 5.1, quantifying the physical effect
of interest, is the value of the estimator, Y, which allows the conclusion that the physical effect is present, if the
primary measurement result, y, exceeds the decision threshold, y* . If the result, y, is below the decision
threshold, y* , the result cannot be attributed to the physical effect, nevertheless it cannot be concluded that it
is absent. If the physical effect is really absent, the probability of taking the wrong decision, that the effect is
present, is equal to the specified probability, α (error of the first kind; see 6.1 and 6.5).
A determined primary measurement result, y, for the non-negative measurand is only significant for the true

value of the measurand to differ from zero ( y > 0 ), if it is larger than the decision threshold

y* = k u(0) (21)
1−α
6.3 Detection limit
#
The detection limit, y , is the smallest true value of the measurand, for which, by applying the decision rule
according to 6.2, the probability of the wrong assumption that the physical effect is absent (error of the second
kind) does not exceed the specified probability, β (see 6.1).
In order to find out whether a measurement procedure is suitable for the measurement purpose, the detection
#
limit, y , is compared with the specified guideline value, y , of the measurand (see 6.1 and 6.6). The
r
#
detection limit, y , is the smallest true value of the measurand which can be detected with the measurement
procedure used. It is high enough compared to the decision threshold, y* , that the probability of the error of
#
the second kind does not exceed β. The detection limit, y , is obtained as the smallest solution of
Equation (22):
##

y =+y* k u()y (22)
1−β
#
where yyW* always applies.
10 © ISO 2010 – All rights reserved

#
Equation (22) is an implicit equation; its right-hand side also depends on y . The detection limit can be
#
calculated by solving Equation (22) for y or, more simply, by iteration (see Figure 1). Thus, an improved
#
 
approximation y is obtained by repeatedly substituting an approximation y for y on the right-hand side
i+1 i
of Equation (22). This leads to Equation (23):

y =+y* k u()y (23)
ii+−11 β

As a starting approximation, for instance, yy= 2* can be chosen. The iteration converges in most cases,
# #
but not if Equation (22) does not have a solution, y . In the latter case or if yy<* results, a detection limit
cannot be established for this measurement procedure (see 5.3.2 and 6.6). In this situation enlarging β can
result in a solution to Equation (22).
 
After the calculation of y or, for instance, with a suitable choice of e.g. yy= 3* , it is more advantageous for
1 1
iW 1 to apply the regula falsi, which in general converges more rapidly. For this purpose, Equation (23) shall
be replaced by
⎡⎤   
y*+⋅k y u()y− y u(y) (y− y)
1−β ij j i i j
⎣⎦
y = (24)
i+1
⎡⎤   
1(−⋅kuy)−u(y)(y−y)
1−β ij i j
⎣⎦
with j < i . Therefore, j = 0 should be set or j be fixed after several iteration steps (see Figure 1).

Key

1 straight line yy=

2 curve yy=+* k u()y
1−β

y true value of the measurand
y estimate of the measurand
For the other symbols, see text.
Figure 1 — Calculation of the detection limit by iteration

With the iteration according to Equations (23) or (24) and beginning with a starting approximation, y , for
instance, yy = 2* as shown, the sequences of the improved approximations y ( j =1, 2,.) converge to
0 i
#
the detection limit y , which is the abscissa of the intersection point of the straight line (key item 1 in
Figure 1) and curve (key item 2 in Figure 1). y* is the decision threshold. With the alternative application of

the regula falsi according to Equation (24), the sequence, y , is generated by means of secants of curve 2, for
i
instance, through points A and B. The shown hyperbolic shape of the curve (key item 2 in Figure 1) is typical
of many applications, for instance, those with Equations (14) or (16). The detection limit does not exist if the
curve does not intersect the straight line (key item 1 in Figure 1) at any abscissa yyW * .
Any iteration shall be stopped if a specified accuracy of ν digits is attained, i.e. if the first digits, ν, of the
successive approximations no longer change. But, if too high an accuracy is demanded, then, even with an
iteration converging in principle, the successive approximations in general permanently fluctuate around and
close to the exact solution but never attain it. A smaller ν shall then be chosen.
#
With the approximation uy( ) =uy( ) (see 5.3.3), y=+()kk u(y) applies.
11−−αβ
The linear interpolation according to Equation (19) leads to the approximation:
#22 22

ya=+a +()k −k u(0) (25)
11−−βα
with
22 2
⎡ ⎤
ak=+u(0) k y (u (y )−u (0) (26)
11−−αβ{()1 1 }
⎣ ⎦
#
If , ya= 2 follows.
α = β
If α = β is chosen and uy() is obtained or approximated by a second-order polynomial
  
uy()=+c cy+cy as in Equations (14), (16) and (20), then with kk==k
11−−αβ
01 2
y* = k c (27)
and
2y* +k c
#
y = (28)
1−kc
6.4 Limits of the confidence interval
The limits of the confidence interval are provided for a physical effect, recognized as present according to 6.2,
in such a way that the confidence interval contains the true value of the measurand with the specified
probability 1− γ (see 6.1). The limits of the confidence interval take into account the fact that the measurand
is non-negative.
With a primary measurement result, y, of the measurand and the standard uncertainty, uy( ) , associated with y
 
(see 5.2), the lower limit of the confidence interval, y , and the upper limit of the confidence interval, y , are
provided by:

y =−yk u()y with p=⋅ωγ(1− 2) (29)
p

y =+yk u()y with q=−12ω γ (30)
q
12 © ISO 2010 – All rights reserved

where
yu/(y)
1 v
ω=−exp( )dvy=Φ /u(y) (31)
⎡⎤
⎣⎦


−∞
For the distribution function, Φ(t) , of the standardized normal distribution and for its inversion, kt= for
p
Φ=()tp , see Table E.1. For methods for its calculation, see Annex E or, for instance, Reference [29].
In general, the limits of the confidence interval are located neither symmetrical to y, nor to the best estimate,
 
yˆ (see 6.5 and Figure 2), but the probabilities of the measurand being smaller than y or larger than y

both equal γ 2 . The relations 0< ω = 1 may be set if yuW 4 (y) . In this case, the following approximations symmetrical to y apply:
 
y =−yk u()y and y =+yk u()y (32)
1/−γ 2 1/−γ 2
6.5 Assessment of a measurement result
The determined primary measurement result, y, of the measurand shall be compared with the decision
threshold, y* . If yy> * , the physical effect quantified by the measurand is recognized as present. Otherwise,
it is decided that the effect is absent.
If yyW * and with ω according to Equation (31), the best estimate yˆ of the measurand is given by (see 5.1
and Figure 2):
⎡⎤
uy()exp −y 2u ()y
{ }
⎣⎦
yyˆ=+ (33)
ωπ2
The standard uncertainty associated with yˆ reads
uy()ˆˆ=−u (y) (y−y)yˆ (34)
NOTE 1 If the best estimate, yˆ, and its standard uncertainty, uy()ˆ, are calculated, the recording of the primary
measurement result, y, and its standard uncertainty, uy(), may be omitted.
NOTE 2 If the decision rule defined by the decision threshold is not used and if yy< *, the best estimate, yˆ, and its
standard uncertainty, uy()ˆ, can also be calculated.

The relations yy< ˆ , 0 < yˆ and yy<<ˆ y as well as uy()ˆ yuW4(y) , the approximations
yyˆˆ== ; u(y) u(y) (35)
hold true.
Key
yuy() quotient of the primary measurement result of the measurand and the standard uncertainty of the measurand
associated with the best estimate
For the curves, see the text.
Figure 2 — Best estimate and limits of the confidence interval

ˆ ˆ
The best estimate, y , of the measurand, the associated standard uncertainty, uy() , the lower limit, y , and

the upper limit, y , of the confidence interval are given as functions of the primary measurement result, y. All
these values are scaled with the standard uncertainty, uy( ) , and γ = 0,05 is chosen. The ascending straight
lines and the horizontal straight line with ordinate 1 are asymptotes.
6.6 Assessment of a measurement procedure
The decision on whether or not a measurement procedure to be applied sufficiently satisfies the requirements
regarding the detection of the physical effect quantified by the measurand is made by comparing the detection
# # #
limit, y , with the specified guideline value, y . If yy> or if Equation (22) has no solution, y , the
r r
measurement procedure is not suitable for the intended measurement purpose with respect to the
requirements.
#
To improve the situation in the case of yy> , it can often be sufficient to choose longer measurement
r
durations or to preselect more counts of the measurement procedure. This reduces the detection limit.
14 © ISO 2010 – All rights reserved

7 Documenta
...


NORME ISO
INTERNATIONALE 11929
Première édition
2010-03-01
Détermination des limites
caractéristiques (seuil de décision, limite
de détection et extrémités de l'intervalle
de confiance) pour mesurages de
rayonnements ionisants — Principes
fondamentaux et applications
Determination of the characteristic limits (decision threshold, detection
limit and limits of the confidence interval) for measurements of ionizing
radiation — Fundamentals and application

Numéro de référence
©
ISO 2010
PDF – Exonération de responsabilité
Le présent fichier PDF peut contenir des polices de caractères intégrées. Conformément aux conditions de licence d'Adobe, ce fichier
peut être imprimé ou visualisé, mais ne doit pas être modifié à moins que l'ordinateur employé à cet effet ne bénéficie d'une licence
autorisant l'utilisation de ces polices et que celles-ci y soient installées. Lors du téléchargement de ce fichier, les parties concernées
acceptent de fait la responsabilité de ne pas enfreindre les conditions de licence d'Adobe. Le Secrétariat central de l'ISO décline toute
responsabilité en la matière.
Adobe est une marque déposée d'Adobe Systems Incorporated.
Les détails relatifs aux produits logiciels utilisés pour la création du présent fichier PDF sont disponibles dans la rubrique General Info
du fichier; les paramètres de création PDF ont été optimisés pour l'impression. Toutes les mesures ont été prises pour garantir
l'exploitation de ce fichier par les comités membres de l'ISO. Dans le cas peu probable où surviendrait un problème d'utilisation,
veuillez en informer le Secrétariat central à l'adresse donnée ci-dessous.

DOCUMENT PROTÉGÉ PAR COPYRIGHT

©  ISO 2010
Droits de reproduction réservés. Sauf prescription différente, aucune partie de cette publication ne peut être reproduite ni utilisée sous
quelque forme que ce soit et par aucun procédé, électronique ou mécanique, y compris la photocopie et les microfilms, sans l'accord écrit
de l'ISO à l'adresse ci-après ou du comité membre de l'ISO dans le pays du demandeur.
ISO copyright office
Case postale 56 • CH-1211 Geneva 20
Tel. + 41 22 749 01 11
Fax + 41 22 749 09 47
E-mail copyright@iso.org
Web www.iso.org
Publié en Suisse
ii © ISO 2010 – Tous droits réservés

Sommaire Page
Avant-propos .iv
Introduction.v
1 Domaine d'application .1
2 Références normatives.2
3 Termes et définitions .2
4 Grandeurs et symboles.4
5 Principes essentiels.6
5.1 Aspects généraux concernant le mesurande.6
5.2 Modèle .6
5.3 Calcul de l'incertitude-type comme fonction du mesurande.8
6 Limites et évaluations caractéristiques .10
6.1 Spécifications .10
6.2 Seuil de décision .10
6.3 Limite de détection.10
6.4 Extrémités de l'intervalle de confiance .12
6.5 Évaluation du résultat de mesurage.13
6.6 Évaluation de la procédure de mesure .14
7 Documentation .15
Annexe A (informative) Vue d'ensemble de la procédure générale.16
Annexe B (normative) Applications diverses .18
Annexe C (normative) Applications à des mesures par comptage spectrométriques.25
Annexe D (informative) Exemples d'application.38
Annexe E (informative) Fonction de distribution de la distribution normale type.50
Annexe F (informative) Notes explicatives.52
Bibliographie.60

Avant-propos
L'ISO (Organisation internationale de normalisation) est une fédération mondiale d'organismes nationaux de
normalisation (comités membres de l'ISO). L'élaboration des Normes internationales est en général confiée
aux comités techniques de l'ISO. Chaque comité membre intéressé par une étude a le droit de faire partie du
comité technique créé à cet effet. Les organisations internationales, gouvernementales et non
gouvernementales, en liaison avec l'ISO participent également aux travaux. L'ISO collabore étroitement avec
la Commission électrotechnique internationale (CEI) en ce qui concerne la normalisation électrotechnique.
Les Normes internationales sont rédigées conformément aux règles données dans les Directives ISO/CEI,
Partie 2.
La tâche principale des comités techniques est d'élaborer les Normes internationales. Les projets de Normes
internationales adoptés par les comités techniques sont soumis aux comités membres pour vote. Leur
publication comme Normes internationales requiert l'approbation de 75 % au moins des comités membres
votants.
L'attention est appelée sur le fait que certains des éléments du présent document peuvent faire l'objet de
droits de propriété intellectuelle ou de droits analogues. L'ISO ne saurait être tenue pour responsable de ne
pas avoir identifié de tels droits de propriété et averti de leur existence.
L'ISO 11929 a été élaborée par le comité technique ISO/TC 85, Énergie nucléaire, sous-comité SC 2,
Radioprotection.
Cette première édition de l'ISO 11929 annule et remplace l'ISO 11929-1:2000, l'ISO 11929-2:2000,
l'ISO 11929-3:2000, l'ISO 11929-4:2001, l'ISO 11929-5:2005, l'ISO 11929-6:2005, l'ISO 11929-7:2005 et
l'ISO 11929-8:2005, qui ont fait l'objet d'une révision technique, en particulier en ce qui concerne le type de
traitement statistique des données.
iv © ISO 2010 – Tous droits réservés

Introduction
Les limites spécifiées dans la présente Norme internationale, censées être fournies au moyen de tests
statistiques et de probabilités spécifiées, permettent d'évaluer les possibilités de détection d'un mesurande
ainsi que l'effet physique quantifié par ce mesurande, comme suit:
⎯ le «seuil de décision» permet de décider si l'effet physique quantifié par le mesurande est présent ou
non;
⎯ la «limite de détection» indique la plus petite valeur vraie du mesurande qui peut encore être détectée par
la procédure de mesurage utilisée; cela permet de décider si la procédure satisfait ou non aux exigences
et si elle est donc adaptée à l'objectif de mesurage prévu;
⎯ les «extrémités de l'intervalle de confiance» comprennent, si l'effet physique est reconnu comme présent,
un intervalle de confiance comprenant la valeur vraie du mesurande avec une probabilité spécifiée.
Les limites mentionnées ci-dessus sont appelées conjointement «limites caractéristiques».
Comme l'incertitude de mesure joue un rôle important dans la présente Norme internationale, l'évaluation des
mesurages et le traitement des incertitudes associées sont réalisés au moyen de procédures générales
conformément au Guide ISO/CEI 98-3; voir également les Références [1, 2]. Cela permet d'établir une
séparation stricte entre l'évaluation des mesurages d'une part (Article 5) et la mise en place et le calcul des
limites caractéristiques d'autre part (Article 6). La présente Norme internationale est basée sur les méthodes
statistiques Bayesiennes, conformément aux Références [6 à 19], afin de pouvoir tenir également compte de
grandeurs incertaines et d'influences qui ne se comportent pas de manière aléatoire lors de mesurages
répétés plusieurs fois ou lors de mesurages par comptage.
Des équations sont fournies pour calculer les limites caractéristiques d'un mesurande des rayonnements
ionisants via «l'incertitude-type de mesure» du mesurande (appelée ci-après «incertitude-type»). Les
incertitudes-types de mesure, ainsi que celles du traitement de l'échantillon, de l'étalonnage du système de
mesure et autres influences, sont prises en compte. Toutefois, ces dernières incertitudes-types sont
supposées être connues grâce à des recherches antérieures.

NORME INTERNATIONALE ISO 11929:2010(F)

Détermination des limites caractéristiques (seuil de décision,
limite de détection et extrémités de l'intervalle de confiance)
pour mesurages de rayonnements ionisants — Principes
fondamentaux et applications
1 Domaine d'application
La présente Norme internationale spécifie une procédure, dans le domaine de la métrologie des
rayonnements ionisants, permettant de calculer le «seuil de décision», la «limite de détection» et les
«extrémités de l'intervalle de confiance» d'un mesurande non négatif caractéristique d'un rayonnement
ionisant, lorsque des mesurages par comptage avec présélection de temps ou du nombre d'impulsions sont
réalisés, et que le mesurande résulte d'un taux de comptage brut et d'un taux de comptage du bruit de fond
ainsi que de grandeurs supplémentaires sur la base d'un modèle d'évaluation. En particulier, le mesurande
peut être le taux de comptage net défini comme étant la différence du taux de comptage brut et du taux de
comptage du bruit de fond, ou l'activité nette d'un échantillon. Il peut également être influencé par l'étalonnage
du système de mesure, par le traitement de l'échantillon et par d'autres facteurs.
La présente Norme internationale s'applique également, de la même manière, à:
⎯ des mesurages par comptage d'objets en mouvement (voir B.2);
⎯ des mesurages réalisés avec des appareils de mesure analogiques de taux de comptage à échelle
linéaire (ci-après appelés ictomètres, voir B.3);
⎯ des mesurages par comptage répétés avec des influences aléatoires (voir B.4);
⎯ des mesurages par comptage sur filtres pendant l'accumulation de matériaux radioactifs (voir B.5);
⎯ des mesurages par comptage spectrométriques multicanal, lorsque des raies particulières du spectre
doivent être prises en considération et si aucun calcul d'ajustement, par exemple la déconvolution, ne
doit être effectué (voir C.2 à C.4);
⎯ des mesurages par comptage spectrométriques multicanal, en cas d'évaluation par des méthodes de
déconvolution (voir C.5), en particulier les mesurages spectrométriques alpha et gamma (voir
respectivement C.5.5 et C.5.6).
La présente Norme internationale s'applique également de manière analogue à d'autres mesurages de tout
type, lorsque le même modèle d'évaluation est concerné. Dans ce sens, elle est également applicable aux
[18]
mesurages réalisés avec des dosimètres albédo . D'autres exemples pratiques sont disponibles dans
[21] [22] [23] [24]
d'autres Normes internationales, par exemple l'ISO 18589 , l'ISO 9696 , l'ISO 9697 , l'ISO 9698 ,
[25] [26] [27] [28]
l'ISO 9699 , l'ISO 10703 , l'ISO 7503 et l'ISO 28218 .
2 Références normatives
Les documents de référence suivants sont indispensables pour l'application du présent document. Pour les
références datées, seule l'édition citée s'applique. Pour les références non datées, la dernière édition du
document de référence s'applique (y compris les éventuels amendements).
ISO 31-0, Grandeurs et unités — Partie 0: Principes généraux
ISO 31-9, Grandeurs et unités — Partie 9: Physique atomique et nucléaire
Guide ISO/CEI 98-3:2008, Incertitude de mesure — Partie 3: Guide pour l'expression de l'incertitude de
mesure (GUM:1995)
Guide ISO/CEI 99:2007, Vocabulaire international de métrologie — Concepts fondamentaux et généraux et
termes associés (VIM)
ISO 3534-1, Statistique — Vocabulaire et symboles — Partie 1: Termes statistiques généraux et termes
utilisés en calcul des probabilités
3 Termes et définitions
Pour les besoins du présent document, les termes et définitions donnés dans l'ISO 31-0, l'ISO 31-9,
le Guide ISO/CEI 98-3, le Guide ISO/CEI 99 et l'ISO 3534-1 ainsi que les suivants s'appliquent.
3.1
procédure de mesure
description détaillée d'un mesurage conformément à un ou plusieurs principes de mesure et à une méthode
de mesure donnée, fondée sur un modèle de mesure et incluant tout calcul destiné à obtenir un résultat de
mesure
[Guide ISO/CEI 99:2007, 2.6]
3.2
mesurande
grandeur que l'on veut mesurer
[Guide ISO/CEI 99:2007, 2.3]
NOTE Dans la présente Norme internationale, un mesurande est non négatif et quantifie un effet de rayonnements
nucléaires. Cet effet n'est pas présent si la valeur vraie du mesurande est zéro. Un exemple de mesurande est l'intensité
de la raie d'un spectre au-delà du bruit de fond dans un mesurage spectrométrique.
3.3
résultat d'un mesurage
ensemble de valeurs attribuées à un mesurande, complété par toute autre information pertinente disponible
[Guide ISO/CEI 99:2007, 2.9]
3.4
incertitude de mesure
incertitude
paramètre non négatif qui caractérise la dispersion des valeurs attribuées à un mesurande, à partir des
informations utilisées
[Guide ISO/CEI 99:2007, 2.26]
Voir également le Guide ISO/CEI 98-3.
NOTE L'incertitude d'un mesurage, dérivée selon le Guide ISO/CEI 98-3, comprend généralement plusieurs
composantes. Certaines peuvent être évaluées grâce à la distribution statistique des résultats de séries de mesurages et
2 © ISO 2010 – Tous droits réservés

peuvent être caractérisées par des écarts-types expérimentaux. Les autres composantes, qui peuvent également être
caractérisées par des écarts-types, sont évaluées à partir de distributions de probabilité supposées ou connues et basées
sur l'expérience ainsi que sur d'autres informations.
3.5
modèle d'évaluation
ensemble de relations mathématiques entre toutes les grandeurs mesurées et les autres grandeurs
impliquées dans l'évaluation de la mesure
NOTE Le modèle d'évaluation n'est pas nécessairement une fonction explicite. Il peut être également un algorithme
réalisé par un code de calcul informatique.
3.6
seuil de décision
valeur de l'estimateur du mesurande telle que, quand le résultat d'une mesure réelle utilisant une procédure
de mesure donnée d'un mesurande quantifiant le phénomène physique lui est supérieur, on décide que le
phénomène physique est présent
NOTE 1 Le seuil de décision est défini de manière que, dans le cas où le résultat du mesurage, y, dépasse le seuil de
décision, y*, la probabilité que la valeur vraie du mesurande soit nulle est inférieure ou égale à la probabilité choisie, α.
NOTE 2 Si le résultat, y, est inférieur au seuil de décision, y*, le résultat ne peut pas être attribué à l'effet physique;
néanmoins, il ne peut être conclu qu'il est absent.
3.7
limite de détection
plus petite valeur vraie du mesurande qui garantit une probabilité spécifiée qu'il soit détectable par la méthode
de mesure
NOTE Avec le seuil de décision conforme à 3.6, la limite de détection est la plus petite valeur vraie du mesurande
pour laquelle la probabilité de décider de façon erronée que la valeur vraie du mesurande est nulle est égale à une valeur
spécifiée, β, quand, en réalité, la valeur vraie du mesurande n'est pas nulle.
3.8
extrémités de l'intervalle de confiance
valeurs qui définissent un intervalle de confiance contenant la valeur vraie du mesurande avec une probabilité
spécifiée
NOTE Un intervalle de confiance est parfois appelé intervalle crédible ou intervalle bayesien. Il est caractérisé dans
la présente Norme internationale pour une probabilité spécifiée (1− γ ).
3.9
meilleur estimateur de la valeur vraie du mesurande
valeur attendue de la distribution des probabilités de la valeur vraie du mesurande, étant donné le résultat
expérimental et toutes les informations obtenues préalablement à la réalisation du mesurage sur le
mesurande
NOTE La meilleure estimation, parmi toutes les estimations possibles du mesurande sur la base des informations
données, qui est associée à l'incertitude minimale.
3.10
valeur de référence
valeur qui correspond aux exigences scientifiques, juridiques ou autres qui est censée être évaluée par la
procédure de mesure
NOTE 1 La valeur de référence peut être donnée, par exemple, comme une activité, une activité spécifique ou une
concentration d'activité, une activité de surface ou un débit de dose.
NOTE 2 La comparaison de la limite de détection avec une valeur de référence permet de déterminer si la procédure
de mesure satisfait ou non aux exigences énoncées par la valeur de référence et de garantir qu'elle est adaptée à
l'objectif du mesurage prévu. La procédure de mesure satisfait à l'exigence si la limite de détection est inférieure à la
valeur de référence.
3.11
bruit de fond
effet de mesurage provoqué par des rayonnements autres que ceux occasionnés par l'objet de mesurage lui-
même
EXEMPLE Sources de radiation naturelles.
3.12
comptage net
contribution due aux rayonnements éventuels d'un objet soumis au mesurage (par exemple d'une source ou
d'un champ de rayonnements) sur l'effet du mesurage
3.13
comptage brut
effet du mesurage provoqué par le bruit de fond et par le comptage net
3.14
facteur d'écran
facteur décrivant la réduction du taux de comptage du bruit de fond par l'effet d'écran provoqué par l'objet du
mesurage
3.15
constante de temps de relaxation
durée pendant laquelle le signal de sortie d'un ictomètre à échelle linéaire diminue pour atteindre 1/e fois la
valeur de départ après l'arrêt de la séquence des impulsions d'entrée
3.16
bruit de fond
〈mesurage spectrométrique〉 nombre d'événements sans intérêt dans la région d'une raie considérée du
spectre
NOTE Ces événements peuvent être dus au bruit de fond provoqué par les rayonnements environnementaux et à
l'échantillon lui-même (par exemple à partir d'autres raies).
4 Grandeurs et symboles
Les symboles des grandeurs auxiliaires et les symboles utilisés uniquement dans les annexes ne sont pas
répertoriés. Les grandeurs physiques sont désignées par des lettres majuscules mais doivent être
soigneusement distinguées de leurs valeurs, désignées par les lettres minuscules correspondantes. Voir
C.5.1 et F.2.1 pour les grandeurs et symboles spéciaux relatifs à la déconvolution dans les mesurages
spectrométriques et aux méthodes statistiques bayesiennes.
m nombre de grandeurs d'entrée
X grandeur d'entrée (1im=,., )
i
x estimateur de la grandeur d'entrée X
i i
ux() incertitude-type de la grandeur d'entrée X associée à l'estimateur x
i i i
hx() incertitude-type ux() , fonction de l'estimateur x
11 1 1
∆x largeur de la région des valeurs possibles de la grandeur d'entrée X
i i
uw() incertitude-type relative de la grandeur d'entrée Y associée à l'estimateur w
rel k
G fonction du modèle
4 © ISO 2010 – Tous droits réservés

Y variable aléatoire, estimateur du mesurande; également utilisé comme symbole du mesurande non
négatif lui-même, qui quantifie le phénomène physique d'intérêt
  
y valeur vraie du mesurande, si l'effet physique d'intérêt n'est pas présent, alors y = 0 , sinon y > 0
y valeur déterminée de l'estimateur Y, estimateur du mesurande, résultat du mesurage primaire du
mesurande
y valeurs y obtenues à partir de différents mesurages ( j = 0,1,2, .)
j
uy() incertitude-type du mesurande associé au résultat du mesurage primaire y
 
uy() incertitude-type de l'estimateur Y, fonction de la valeur vraie y du mesurande
yˆ meilleur estimateur du mesurande
ˆ ˆ
uy() incertitude-type du mesurande associé au meilleur estimateur y
y* seuil de décision du mesurande
#
y limite de détection du mesurande
#

y approximations de la limite de détection y
i
y valeur de référence du mesurande
r
 
y , y limite basse et haute, respectivement, de l'intervalle de confiance du mesurande
ρ taux de comptage comme grandeur d'entrée X
i
i
ρ taux de comptage net
n
ρ taux de comptage brut
g
ρ taux de comptage du bruit de fond
n nombre d'impulsions sommées obtenues à partir du mesurage du taux de comptage ρ
i i
n , n nombre d'impulsions sommées du comptage brut et du comptage du bruit de fond, respectivement
g 0
t durée du mesurage du taux de comptage ρ
i i
t , t durée du mesurage de l'effet brut et de l'effet du bruit de fond, respectivement
g 0
r estimateur du taux de comptage ρ
i i
r , r estimateur du taux de comptage brut et du taux de comptage du bruit de fond, respectivement
g 0
τ , τ constante de temps de relaxation d'un ictomètre utilisé pour le mesurage du taux de comptage brut
g 0
et du taux de comptage du bruit de fond, respectivement
α, β probabilités de l'erreur de première et de deuxième espèces, respectivement
1− γ probabilité relative à l'intervalle de confiance du mesurande
k , k quantiles d'une distribution normale type pour les probabilités p et q, respectivement (par exemple
p q
p=−1 α , 1− β ou 12− γ )
Φ()t fonction de distribution de la distribution normale type; Φ=()kp s'applique
p
5 Principes essentiels
5.1 Aspects généraux concernant le mesurande
Un mesurande non négatif doit être attribué à l'effet physique à étudier pour un mesurage donné. Le

mesurande doit quantifier l'effet. On considère que la valeur vraie est y = 0 , si l'effet n'est pas présent dans
un cas particulier.
Soit une variable aléatoire Y, un estimateur est attribué au mesurande. Le symbole Y est également utilisé
ci-après pour le mesurande lui-même. La valeur y de l'estimateur Y, déterminée à partir de mesurages, est
une estimation du mesurande. Elle doit être calculée comme résultat du mesurage primaire, avec l'incertitude-
type primaire, uy( ) , du mesurande associé à y. Ces deux valeurs forment le résultat complet du mesurage
primaire du mesurande et sont obtenues conformément au Guide ISO/CEI 98-3 (voir également les
Références [1, 2]) par l'évaluation des données de mesurage ou d'autres informations obtenues au moyen
d'un modèle (d'évaluation) reliant mathématiquement toutes les grandeurs concernées (voir 5.2). En général,
le fait que le mesurande soit non négatif n'est pas explicitement pris en compte dans l'évaluation. Par

conséquent, y peut être négatif, surtout lorsque le mesurande s'approche de la valeur vraie y = 0 . Le résultat
de mesure primaire, y, diffère de la meilleure estimation, yˆ , du mesurande, calculée en 6.5. Avec yˆ , le fait
ˆ ˆ
que le mesurande ne puisse être négatif est pris en compte. L'incertitude-type, uy() , associée à y est
inférieure à uy( ) .
5.2 Modèle
5.2.1 Modèle général
Dans de nombreux cas, le mesurande, Y, est fonction de plusieurs grandeurs d'entrée, X , selon l'Équation (1):
i
YG= (X , .,X ) (1)
1 m
L'Équation (1) est le modèle d'évaluation. En substituant les estimateurs, x , aux grandeurs d'entrée, X , dans
i i
la fonction modèle, G, l'Équation (1) devient le résultat de mesurage primaire, y, du mesurande:
yG= (x , .,x ) (2)
1 m
L'incertitude-type, (uy) , du mesurande associée au résultat du mesurage primaire, y, est calculée selon
l'Équation (3) si les grandeurs d'entrée, X, sont mesurées de manière indépendante et si les
i
incertitudes-types, (ux) , associées aux estimateurs, x , sont données:
i i
m
⎛⎞
∂G
uy() = u (x) (3)
⎜⎟
∑ i
∂X
⎝⎠i
i=1
Dans l'Équation (3), les estimateurs, x, doivent être remplacés par les grandeurs d'entrée, X, dans les
i i
dérivées partielles de G. La détermination des estimateurs, x, et des incertitudes-types associées, ux( ) ,
i i
ainsi que la détermination numérique ou expérimentale des dérivées partielles, s'appuient sur le Guide
ISO/CEI 98-3 ou les Références [1, 2]. Pour un taux de comptage, X = ρ , avec le résultat de comptage
ii
donné, n, enregistré pendant une durée de mesurage, t, les spécifications, xr==nt et
i i ii i i
ux()==n t r t, s'appliquent (voir également F.1). Lorsque les grandeurs d'entrée ne sont pas
iii ii
mesurées de manière indépendante et pour des évaluations de mesure plus complexes telles que la
déconvolution, voir C.5.2.
En 5.2.2, la grandeur d'entrée, X , par exemple le taux de comptage brut, est considérée comme une

grandeur dont la valeur, x , n'est pas donnée lorsqu'une valeur vraie, y , du mesurande, Y, est spécifiée dans
i
le cadre du calcul du seuil de décision et de la limite de détection. De façon analogue, la grandeur d'entrée,
X , est attribuée de manière appropriée à l'effet du bruit de fond. Les données relatives aux autres grandeurs
d'entrée sont considérées comme étant des données obtenues à partir de recherches antérieures menées de
manière indépendante.
6 © ISO 2010 – Tous droits réservés

5.2.2 Modèle de mesurage des rayonnements ionisants

Dans la présente Norme internationale, le mesurande, Y, de valeur vraie, y , caractéristique d'un échantillon
de matériaux radioactifs doit être déterminé à partir du comptage de l'effet brut et du comptage de l'effet du
bruit de fond obtenus avec une présélection du temps ou du nombre d'impulsions. En particulier, Y peut être
le taux de comptage net, ρ , ou l'activité nette, A, de l'échantillon. Les symboles relatifs au comptage de l'effet
n
brut et de l'effet du bruit de fond sont indiqués ci-après respectivement par les indices g et 0.
Dans la présente Norme internationale, le modèle est spécifié comme suit:
XX ⋅⋅⋅
Y G(X , .,X ) (XXX X ) (XXX X )W
==−−⋅ =−−⋅ (4)
11m 234 1234
XX ⋅⋅⋅
avec
XX ⋅⋅⋅
W = (5)
XX ⋅⋅⋅
X = ρ est le taux de comptage brut et X = ρ est le taux de comptage du bruit de fond. Les autres
1g 20
grandeurs d'entrée X sont l'étalonnage, les grandeurs de correction ou d'influence, ou les facteurs de
i
conversion, par exemple la probabilité d'émission ou de réponse ou, en particulier, X est un facteur de
protection et X une grandeur supplémentaire de correction du bruit de fond. Si certaines grandeurs d'entrée
ne sont pas concernées, alors x =1 ii=>3; 4 , x = 0 et ux( ) = 0 seront fixés. Pour les taux de
()
i 4 i
comptage,xr==n t et ux()==n t r t ainsi que xr==n t et ux()==n t r t
1g g g 1gg gg 20 0 0 200 00
s'appliquent.
En introduisant les estimateurs, x , dans l'Équation (4), l'estimation primaire, y, du mesurande, Y, donne le
i
résultat suivant:
⎛⎞
n
n
g
y==G(x , .,x ) (x−xxx− )⋅w= (r−rx−x )⋅w=⎜⎟− xx−⋅w (6)
1 m 1 23 4 g 03 4 3 4
⎜⎟
tt
g0
⎝⎠
avec
xx ⋅⋅⋅
w = (7)
xx ⋅⋅⋅
Avec les dérivées partielles:
∂G ∂G ∂G ∂G ∂GY
= W ; =−XW ; =−XW ; =−W ; =± (5iW), (8)
3 2
∂X ∂X ∂X ∂X ∂XX
1 2 3 4 ii
et, en introduisant les estimateurs, x , w et y, l'Équation (3) conduit à l'incertitude-type uy( ) du mesurande
i
associé à y:
2 2 22 22 2 2 2
⎡⎤
uy()= w⋅ u(x)+++x u()x x u()x u(x)+y u (w)
13 2 2 3 4 rel
⎣⎦
(9)
22 22 2 22
⎡⎤
=⋅wr t+xr t+ru()x+u(x)+yu (w)
gg 30 0 0 3 4 rel
⎣⎦

m 2
ux()
i
uw() = (10)
rel ∑
x
i=5 i
qui est la somme des carrés des incertitudes-types relatives des grandeurs X à X . Pour m < 5, les valeurs
5 m
w = 1 et uw() =0 s'appliquent.
rel
L'estimateur, x, et l'incertitude-type, ux( ) , de X (im= 3,., ) sont considérés comme connus grâce aux
i i i
recherches antérieures ou comme des valeurs fondées sur l'expérience provenant d'autres informations: x
i
peut être déterminé comme une valeur arithmétique moyenne et ux() comme une variance empirique
i
(voir B.4.1). Si nécessaire, ux() peut également être calculé comme la variance d'une distribution
i
rectangulaire des valeurs possibles de X d'étendue ∆x . Cela conduit à ux()=∆(x) 12.
i i ii
Pour l'application de la méthode à des mesurages particuliers, y compris les mesures spectrométriques, voir
les Annexes B et C.
5.3 Calcul de l'incertitude-type comme fonction du mesurande
5.3.1 Aspects généraux
Pour la mise en place et le calcul numérique du seuil de décision, en 6.2, et de la limite de détection, en 6.3,
 
l'incertitude-type du mesurande est nécessaire en tant que fonction uy( ) de la valeur vraie, yW0, du
mesurande. Cette fonction doit être déterminée de la même manière que uy( ) dans le cadre de l'évaluation
des mesurages conformément au Guide ISO/CEI 98-3; voir également les Références [1, 2]. Dans la plupart
des cas, uy( ) doit être construite comme une racine carrée positive de la fonction de variance, uy() ,

calculée préalablement. Cette fonction doit être définie, unique et continue pour tous yW0 et ne doit pas
adopter de valeurs négatives.

Dans la plupart des cas, uy( ) peut être spécifié de manière explicite, à condition que ux() soit donné
comme fonction hx() de x . Dans de tels cas, y doit être formellement remplacé par y et l'Équation (2) doit
11 1

être résolue pour x . Avec un y spécifié, la valeur x peut également être calculée de manière numérique à
1 1
partir de l'Équation (2); par exemple, grâce à une méthode d'itération, par laquelle on obtient x comme

fonction de y et x , ., x . Cette fonction doit remplacer x dans l'Équation (3) et dans ux() =h(x), qui
2 m 1
donne finalement uy( ) et non uy( ) . Dans le cas du modèle donné par l'Équation (6) et en 5.3.2, il est

nécessaire de procéder de cette manière. Sinon, 5.3.3 doit être appliqué, où uy( ) suit une approximation par
interpolation à partir des données, y et uy( ) , de plusieurs mesurages.
i j
5.3.2 Calcul explicite
Lorsque, dans le cas du modèle donné par l'Équation (6), l'incertitude-type, ux() , du taux de comptage brut,
X = ρ , est donnée comme fonction hx() de l'estimateur, xr= , alors hx() = x t ou
1g 11 1g 11 1 g
hx() =x n s'applique, si la durée de mesurage, t (présélection du temps), ou, respectivement, le
11 1 g g
nombre d'impulsions enregistrées, n (présélection du nombre d'impulsions), sont spécifiés.
g

La valeur y doit être formellement remplacée par y . Cela permet d'éliminer x dans le cas général, n en
1 g
mode présélection de temps, et t en mode présélection du nombre d'impulsions, dans l'Équation (9) en
g

utilisant l'Équation (6). Ces valeurs ne sont pas disponibles lorsque y est spécifié. Selon l'Équation (6), cela
donne, dans le cas général:
xy=+w xx+x (11)
En remplaçant x selon l'Équation (11) dans la fonction donnée, hx(), c'est-à-dire avec
1 11
ux()=+h(yw xx+x ) , on obtient, à partir de l'Équation (9), l'équation suivante:
11 23 4
22 22 22 2 22
⎡⎤
uy()= w⋅ h(y w+x x+x)+++x u()x x u()x u(x)+y u (w) (12)
12343 22 3 4 rel
⎣⎦
8 © ISO 2010 – Tous droits réservés

Avec une présélection de temps et parce que xn= t et xr= ,
1g g 20

nt=⋅()yw+rx +x (13)
gg 03 4
est obtenu à partir de l'Équation (11). Puis, avec hx()==x t n t et en remplaçant n selon
11 1 g g g g
l'Équation (13) et avec ux() =r t , l'Équation (12) conduit à
⎡⎤
  
uy()=⋅w (y w+r x+x ) t+x r t+r u (x )+u (x )+y u (w) (14)
03 4 g 3 0 0 0 3 4 rel
⎣⎦
Avec une présélection du nombre d'impulsions,
n
g
t = (15)
g

yw++r x x
03 4
est obtenu de manière analogue. Puis, avec hx()==x t n t et en remplaçant t selon l'Équation (15)
11 1 g g g g
et avec ux() =r n , l'Équation (12) conduit à
20 0
⎡⎤
uy()=⋅w (y/w+r x+x ) n+x r n+r u (x )+u (x )+y u (w) (16)
03 4 g 3 0 0 0 3 4 rel
⎣⎦
#
L'Équation (22) a une solution, qui est la limite de détection, y , si, avec une présélection de temps, la
condition suivante est satisfaite:
ku ()w <1 (17)
1r−β el
ou bien, avec une présélection du nombre d'impulsions, la condition suivante est satisfaite:
ku⋅+ ()w<1 (18)
1−β
rel
n
g
Dans le cas contraire, il se peut qu'aucune limite de détection n'existe à cause d'une incertitude trop
importante des grandeurs X à X , exprimée sommairement par uw(). Selon l'Équation (17) la condition
5 m rel
s'applique également dans le cas de l'Équation (12) si hx() augmente pour x croissant plus lentement que
x , c'est-à-dire si hx()x →0 pour x →∞ .
1 11 1 1
5.3.3 Approximations

Il suffit souvent d'utiliser les approximations suivantes pour la fonction uy( ) , en particulier si l'incertitude-type,
ux() , n'est pas connue comme fonction hx() . Une condition préalable est que les résultats de mesurage, y ,
1 11
j
et les incertitudes-types associées, uy( ), calculées selon 5.1 et 5.2, à partir de précédents mesurages du
j
même type, doivent déjà être disponibles ( j = 0,1,2,.). Les mesurages doivent être effectués sur différents
échantillons d'activités différentes mais, à d'autres égards, dans des conditions, autant que possible,
identiques. Un des mesurages peut être un mesurage de l'effet du bruit de fond ou d'un mesurage du blanc
 
avec 0y = et, par exemple, j = 0. Puis, y = 0 doit être déterminé et uu(0) = (y ) . Le mesurage en cours
0 0
de réalisation peut être pris comme un mesurage supplémentaire avec j = 1.
 
La fonction uy( ) présente souvent une croissance plutôt lente. Par conséquent, l'approximation uy() =uy( )
suffit dans certains de ces cas, surtout si le résultat de mesure primaire, y , du mesurande n'est pas
beaucoup plus grand que l'incertitude-type associée, uy() .

Si seules uu(0) = (y ) , y > 0 et uy() sont connues, alors l'interpolation linéaire suivante suffit souvent:
0 1 1
22 2
   
uy()=−u (0)(1 yy )+u (y )yy (19)
11 1
Si les résultats y , y et y ainsi que les incertitudes types associées uy() , uy() et uy() obtenus à partir
0 1 2
0 1 2
de trois mesurages sont disponibles, alors l'interpolation bilinéaire suivante peut être utilisée:
  
()yy−−(yy) ()yy−−()yy (y−y)(y−y)
22 2 2
12 02 0 1

uy()=⋅uy() +uy( )⋅ +uy()⋅ (20)
01 2
(yy−−)(yy) ()y−−y()y y (yy−)(yy−)
01 0 2 1 0 1 2 2 0 2 1
Si les résultats issus de nombreux mesurages identiques sont donnés, alors la forme parabolique de la
fonction uy() peut également être déterminée par un calcul d'ajustement.
6 Limites et évaluations caractéristiques
6.1 Spécifications
La probabilité, α, de l'erreur de première espèce, la probabilité, β, de l'erreur de seconde espèce et la
probabilité, 1− γ , relative à l'intervalle de confiance doivent être spécifiées. Le choix dépend de l'application.
Un choix fréquemment cité est α = β = 0,05. Alors kk== 1, 65 . Si une valeur de 0,05 est choisie pour γ,
11−−αβ
alors k = 1, 96 (voir l'Annexe E).
12−γ
Lorsque l'on doit évaluer si une procédure de mesure satisfait aux exigences à remplir pour des raisons
scientifiques, légales ou autres (voir 6.6), alors une valeur de référence, y , doit également être spécifiée
r
comme valeur du mesurande, par exemple une activité.
6.2 Seuil de décision
Le seuil de décision, y*, du mesurande non négatif selon 5.1, quantifiant l'effet physique d'intérêt, correspond
à la valeur de l'estimateur, Y, qui permet de conclure que l'effet physique est présent, si le résultat de mesure
primaire, y, dépasse le seuil de décision, y*. Si le résultat, y, est inférieur au seuil de décision, y*, le résultat
ne peut être attribué à l'effet physique, néanmoins il ne peut pas être conclu qu'il est absent. Si l'effet
physique est réellement absent, alors la probabilité de prendre la mauvaise décision, à savoir que l'effet est
présent, est égal à la probabilité spécifiée, α (erreur de première espèce; voir 6.1 et 6.5).
Un résultat de mesure primaire, y, déterminé pour le mesurande non négatif n'est significatif, de sorte que la

valeur vraie du mesurande diffère de zéro ( y > 0 ), que s'il est supérieur au seuil de décision.

y* = k u(0) (21)
1−α
6.3 Limite de détection
#
La limite de détection, y , est la plus petite valeur vraie du mesurande pour laquelle, en appliquant la règle
de décision conformément à 6.2, la probabilité de décider à tort que l'effet physique est absent (erreur de
seconde espèce) ne dépasse pas la probabilité spécifiée, β (voir 6.1).
Afin de déterminer si une procédure de mesure est adaptée à l'objectif de mesurage, la limite de détection,
#
y , est comparée à la valeur de référence spécifiée, y , du mesurande (voir 6.1 et 6.6). La limite de détection,
r
#
y , est la plus petite valeur vraie du mesurande pouvant être détectée à l'aide de la procédure de mesure
utilisée. Elle est suffisamment élevée par rapport au seuil de décision, y* , pour que la probabilité d'erreur de
#
seconde espèce ne dépasse pas β. La limite de détection, y , est obtenue comme étant la plus petite valeur
solution de l'Équation (22):
##

y =+y* k u()y (22)
1−β
#
où yyW* s'applique toujours.
10 © ISO 2010 – Tous droits réservés

#
L'Équation (22) est une équation implicite, les termes situés à droite de celle-ci dépendent également de y .
#
La limite de détection peut être calculée en résolvant l'Équation (22) pour y ou, plus simplement, par
#

itération (voir Figure 1). Ainsi, en remplaçant à plusieurs reprises une approximation, y , par y , dans la
i

partie droite de l'Équation (22), on obtient une approximation améliorée, y , selon l'Équation (23):
i+1

y =+y* k u()y (23)
ii+−11 β

Par exemple, yy= 2* peut être choisie comme approximation de départ. L'itération converge dans la
# #
plupart des cas, sauf si l'Équation (22) n'a pas de solution, y . Dans ce dernier cas, ou si yy<* , la limite
de détection ne peut pas être déterminée pour cette procédure de mesure (voir 5.3.2 et 6.6). Dans ce cas,
l'augmentation de β peut conduire à déterminer une solution à l'Équation (22).
 
Après le calcul de y ou, par exemple, avec un choix adapté de yy= 3* , il est plus avantageux pour iW1
1 1
d'appliquer la méthode de regula falsi, qui converge généralement plus rapidement. À cet effet, l'Équation (23)
doit être remplacée par
⎡⎤   
y*+⋅k y u()y− y u(y) (y− y)
1−β ij j i i j
⎣⎦
y = (24)
i+1
⎡⎤   
1(−⋅kuy)−u(y)(y−y)
1−β ij i j
⎣⎦
avec . Il convient ensuite de poser j = 0 ou de fixer j après plusieurs étapes d'itération (voir Figure 1).
j < i
Légende

1 droite yy=


2 courbe yy=+k uy
()
1−β

y valeur vraie du mesurande
y estimateur du mesurande
Pour les autres symboles, voir texte.
Figure 1 — Calcul de la limite de détection par itération
En utilisant la méthode d'itération selon les Équations (23) ou (24) et en commençant par une approximation
de départ, y , par exemple, yy = 2* , les séquences des approximations améliorées y(1j =,2,.)
0 0 i
#
convergent vers la limite de détection, y , qui est l'abscisse du point d'intersection de la droite (repère 1 dans
la Figure 1) et de la courbe (repère 2 dans la Figure 1). y* est le seuil de décision. Grâce à l'application

alternative de la méthode regula falsi selon l'Équation (24), la séquence, y , est générée au moyen de droites
i
sécantes de la courbe (repère 2 dans la Figure 1), par exemple passant par les points A et B. La forme
hyperbolique de la courbe est typique de nombreuses applications, comme celles obtenues par les
Équations (14) ou (16). La limite de détection n'existe pas si la courbe ne coupe pas la droite (repère 1 dans

la Figure 1) pour n'importe quelle abscisse, yyW * .
Toute itération doit être interrompue si une précision spécifiée de ν chiffres est atteinte, c'est-à-dire si les
premiers chiffres, ν, des approximations successives ne changent plus. Toutefois, si une précision trop élevée
est demandée, les approximations successives varient généralement de façon permanente autour et à
proximité de la solution exacte mais sans jamais l'atteindre, même avec une itération en principe convergente.
Une valeur ν inférieure doit alors être choisie.
#

Avec l'approximation uy( ) =uy( ) (voir 5.3.3), y=+()kk u(y) s'applique.
11−−αβ
L'interpolation linéaire selon l'Équation (19) mène à l'approximation:
#22 22

ya=+a +()k −k u(0) (25)
−−βα
avec
22 2
⎡ ⎤
ak=+u(0) k y (u (y )−u (0) (26)
11−−αβ{()1 1 }
⎣ ⎦
#
Si α = β , alors ya= 2 s'ensuit.

Si on choisit α = β et soit on obtient uy() , soit il est approché par un polynôme de second ordre,
uy()=+c cy+cy , comme dans les Équations (14), (16) et (20), alors, avec kk==k :
01 2 11−−αβ
y* = k c (27)
et
2y* +k c
#
y = (28)
1−kc
6.4 Extrémités de l'intervalle de confiance
Les extrémités de l'intervalle de confiance sont données pour un effet physique reconnu comme étant présent
selon 6.2, de manière que l'intervalle de confiance contienne la valeur vraie du mesurande avec la probabilité
spécifiée 1 − γ (voir 6.1). Les extrémités de l'intervalle de confiance prennent en compte le fait que le
mesurande est non négatif.
En utilisant un résultat de mesure primaire, y, du mesurande, et l'incertitude-type, uy( ) , associée à y (voir 5.2),
 
la limite inférieure de l'intervalle de confiance, y , et la limite supérieure de l'intervalle de confiance, y , sont
données par:

y =−yk u()y avec p=⋅ωγ(1− 2) (29)
p

y =+yk u()y avec q=−12ω γ (30)
q
12 © ISO 2010 – Tous droits réservés

où:
yu/(y)
1 v
ω=−exp( )dvy=Φ /u(y) (31)
⎡⎤
⎣⎦


−∞
Pour la fonction de distribution, Φ(t), de la distribution normale réduite, et pour son inversion, kt= , pour
p
Φ=()tp, voir le Tableau E.1. Pour les méthodes de calcul, voir l'Annexe E, ou, par exemple, la
Référence [29].
En général, les extrémités de l'intervalle de confiance ne sont symétriques ni par rapport à y, ni par rapport au

meilleur estimateur, yˆ (voir 6.5 et Figure 2), mais les probabilités que le mesurande soit inférieur à y ou
 
supérieur à y sont toutes les deux égales à γ 2 . Les relations 0< ω = 1 peut être fixé si yuW 4 (y) . Dans ce cas, les approximations suivantes symétriques à y s'appliquent:
 
y =−yk u()y et y =+yk u()y (32)
1/−γ 2 1/−γ 2
6.5 Évaluation du résultat de mesurage
Le résultat du mesurage primaire déterminé, y, du mesurande doit être comparé au seuil de décision, y* . Si
yy> * , alors l'effet physique quantifié par le mesurande est reconnu comme présent. Dans le cas contraire,
on considèr
...

Questions, Comments and Discussion

Ask us and Technical Secretary will try to provide an answer. You can facilitate discussion about the standard in here.

Loading comments...