ISO 19003:2006
(Main)Rubber and rubber products - Guidance on the application of statistics to physical testing
Rubber and rubber products - Guidance on the application of statistics to physical testing
ISO 19003:2006 provides guidance on the application of statistics to rubber testing. It is intended not to conflict with or replace existing International Standards covering basic statistical techniques, but rather to complement them and provide examples of those techniques applied to particular rubber testing situations.
Caoutchouc et produits à base de caoutchouc — Lignes directrices pour l'application des statistiques aux essais physiques
L'ISO 19003:2006 fournit des lignes directrices sur l'application des statistiques aux essais du caoutchouc. Elle est destinée à éviter tout conflit avec des Normes internationales existantes traitant de techniques statistiques de base et ne vise en aucun cas à se substituer à ces normes, son unique objectif étant de compléter ces dernières et d'illustrer par des exemples les techniques appliquées à des situations d'essai particulières du caoutchouc.
General Information
- Status
- Published
- Publication Date
- 03-Oct-2006
- Technical Committee
- ISO/TC 45/SC 2 - Testing and analysis
- Drafting Committee
- ISO/TC 45/SC 2/WG 4 - Application of statistical methods
- Current Stage
- 9093 - International Standard confirmed
- Start Date
- 21-Jan-2025
- Completion Date
- 13-Dec-2025
Overview
ISO 19003:2006 - Rubber and rubber products: Guidance on the application of statistics to physical testing provides practical guidance on applying statistical methods to rubber testing data. It is intended to complement, not replace, existing international standards on basic statistical techniques. The standard gives examples and methodology tailored to typical rubber testing situations, helping laboratories and manufacturers interpret test results, demonstrate conformity, and improve test reliability.
Key topics
The standard covers a broad range of statistical topics applied to rubber physical testing, including:
- Variability, accuracy, trueness and precision - understanding limits of test results and sources of variation.
- Distribution and central tendency - types of distribution, mean/median/mode, measures of dispersion, transforms and tests for normality.
- Confidence limits and significant difference - confidence intervals, comparison of results, specification limits.
- Ranking methods - non-parametric comparisons (e.g., Friedman’s test, outside count test).
- Outlier criteria - guidance and tests for rejecting outliers (Dixon’s test, Cochran’s test for variance).
- Analysis of variance (ANOVA) - one- and multi-factor ANOVA applications to rubber tests.
- Regression analysis - linear, quadratic, cubic fits and practical examples (e.g., temperature or ageing effects).
- Uncertainty of measurement - compiling random and systematic components and reporting uncertainty.
- Sampling and number of test pieces - sampling plans, acceptable quality levels, random sampling and sample sizes.
- Expression of results and precision statements - reporting formats, rounding, histograms.
- Design of experiments - principles and methodology for planning experiments relevant to rubber testing.
Practical applications and users
ISO 19003 is directly useful for:
- Materials testing laboratories performing rubber physical tests (tensile, fatigue, ageing, compression set).
- Quality control and QA teams in rubber product manufacturing, to set specification limits and interpret batch test data.
- R&D engineers and test method developers designing experiments and regression models for material behaviour.
- Accreditation bodies and auditors assessing statistical treatment of test data and uncertainty reporting.
Benefits include more robust interpretation of test data, consistent reporting, better decision-making on pass/fail status, and clearer uncertainty statements.
Related standards
ISO 19003 is intended to complement existing international statistical and metrology standards that provide basic statistical techniques and guidance on measurement uncertainty; users should apply ISO 19003 together with those general statistical references when developing laboratory procedures and reports.
Keywords: ISO 19003, rubber testing, statistics, physical testing, measurement uncertainty, ANOVA, regression, sampling, outliers, design of experiments.
ISO 19003:2006 - Rubber and rubber products — Guidance on the application of statistics to physical testing Released:10/4/2006
ISO 19003:2006 - Caoutchouc et produits à base de caoutchouc — Lignes directrices pour l'application des statistiques aux essais physiques Released:10/4/2006
Frequently Asked Questions
ISO 19003:2006 is a standard published by the International Organization for Standardization (ISO). Its full title is "Rubber and rubber products - Guidance on the application of statistics to physical testing". This standard covers: ISO 19003:2006 provides guidance on the application of statistics to rubber testing. It is intended not to conflict with or replace existing International Standards covering basic statistical techniques, but rather to complement them and provide examples of those techniques applied to particular rubber testing situations.
ISO 19003:2006 provides guidance on the application of statistics to rubber testing. It is intended not to conflict with or replace existing International Standards covering basic statistical techniques, but rather to complement them and provide examples of those techniques applied to particular rubber testing situations.
ISO 19003:2006 is classified under the following ICS (International Classification for Standards) categories: 83.060 - Rubber. The ICS classification helps identify the subject area and facilitates finding related standards.
You can purchase ISO 19003:2006 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 19003
First edition
2006-10-01
Rubber and rubber products — Guidance
on the application of statistics to physical
testing
Caoutchouc et produits à base de caoutchouc — Lignes directrices
pour l'application des statistiques aux essais physiques
Reference number
©
ISO 2006
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 2006
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 2006 – All rights reserved
Contents Page
Foreword. vi
Introduction . vii
1 Scope . 1
2 References. 1
3 Terms and definitions. 1
4 Symbols . 4
5 Limitations of test results . 5
5.1 Variability. 5
5.2 Accuracy, trueness and precision . 6
5.3 Relevance and significance. 7
6 Distribution of results and measures of central tendency. 7
6.1 Principles. 7
6.2 Methodology. 8
6.2.1 Types of distribution . 8
6.2.2 Measures of central tendency . 12
6.2.3 Measures of dispersion. 15
6.2.4 Transformation to normal distribution . 16
6.2.5 Test of departure from normality . 17
6.3 Applications to rubber testing. 20
6.3.1 General. 20
6.3.2 Tensile testing. 20
6.3.3 Fatigue . 21
6.3.4 Conversion to normal distribution. 24
6.3.5 Other uses of the median. 25
7 Confidence limits and significant difference . 26
7.1 Principles. 26
7.2 Methodology. 26
7.2.1 Confidence limits and confidence intervals . 26
7.2.2 Significant difference . 31
7.3 Applications to rubber testing. 35
7.3.1 General. 35
7.3.2 Confidence limits and specification limits. 35
7.3.3 Comparison of results. 36
8 Ranking methods. 37
8.1 Principles. 37
8.2 Methodology. 37
8.2.1 Friedman’s test . 37
8.2.2 The outside count test. 39
8.3 Applications to rubber testing. 39
9 Criteria for rejecting outliers . 40
9.1 Principles. 40
9.2 Methodology. 41
9.2.1 General. 41
9.2.2 Dixon’s test. 41
9.2.3 Cochran’s test for variance . 43
9.3 Applications to rubber testing. 45
9.3.1 General. 45
9.3.2 Dixon’s test applied to individual results . 45
9.3.3 Cochran’s variance test. 46
9.3.4 Dixon’s test applied to a group of mean values . 47
10 Analysis of variance (ANOVA). 47
10.1 Principles . 47
10.2 Methodology. 48
10.2.1 General . 48
10.2.2 One factor with an equal number of replicates. 48
10.2.3 One factor with a variable number of replicates. 49
10.2.4 Two (and over) factor analysis of variance .50
10.3 Applications to rubber testing. 50
11 Regression analysis . 54
11.1 Principles . 54
11.2 Methodology. 55
11.2.1 General . 55
11.2.2 Linear least squares . 55
11.2.3 Quadratic least squares . 56
11.2.4 Cubic least squares . 56
11.3 Applications to rubber testing. 56
11.3.1 General . 56
11.3.2 The effect of temperature on compression set. 56
11.3.3 Effect of ageing on tensile strength. 58
11.3.4 Temperature of retraction test. 59
12 Uncertainty of measurement. 60
12.1 Principles . 60
12.2 Methodology. 61
12.2.1 Compilation of a single value for uncertainty. 61
12.2.2 Random uncertainty (U ) . 61
r
12.2.3 Systematic uncertainty (U ) . 62
s
12.2.4 Deviation of a single value of total uncertainty . 63
12.2.5 Reporting of results . 63
12.3 Applications to rubber testing. 63
13 Sampling . 64
13.1 Principles . 64
13.2 Methodology. 65
13.2.1 General . 65
13.2.2 Acceptable quality level and limiting quality .65
13.2.3 Assessment of nonconformity . 65
13.2.4 Inspection levels . 66
13.2.5 Plans for sampling by attributes . 66
13.2.6 Random sampling . 67
13.3 Applications to rubber testing. 68
14 Number of test pieces. 68
14.1 Principles . 68
14.2 Methodology. 69
14.3 Applications to rubber testing. 69
14.3.1 General . 69
14.3.2 Refinement of confidence limits. 69
14.3.3 Refinement of a pass/fail status . 70
15 Expression of results. 70
15.1 Principles . 70
15.2 Methodology. 70
15.2.1 The test report . 70
15.2.2 Rounding. 72
15.3 Applications to rubber testing. 73
15.3.1 General . 73
15.3.2 Construction of a histogram. 73
iv © ISO 2006 – All rights reserved
15.3.3 Examples of rounding . 74
16 Precision statements. 74
16.1 General. 74
16.2 Principles. 74
16.3 Methodology. 75
16.4 Applications to rubber testing. 77
17 Design of experiments . 78
17.1 General information and principles . 78
17.1.1 General information. 78
17.1.2 Principles. 79
17.2 Methodology. 92
17.2.1 General. 92
17.2.2 Descriptive experiments . 92
17.2.3 Comparative experiments. 93
17.2.4 Response experiments. 95
17.3 Applications to rubber testing. 95
17.3.1 Descriptive experiments . 95
17.3.2 Comparative experiments. 97
17.3.3 Response experiments. 101
18 Statistical quality control . 105
18.1 Principles. 105
18.2 Methodology. 106
18.2.1 General. 106
18.2.2 Control charts by attributes. 106
18.2.3 Control charts by variables . 106
18.3 Applications to rubber testing. 108
18.3.1 General. 108
18.3.2 Control charts . 108
18.3.3 Cusum chart . 111
Annex A (informative) Mathematical form of the distribution functions referenced in this
International Standard. 115
Annex B (informative) Additional forms of mean value . 117
Annex C (informative) Inter-relationships for measures of central tendency in the double
exponential and Weibull distributions. 118
Annex D (informative) Equation for the calculation of standard deviation . 119
Annex E (informative) Construction of Weibull probability paper . 121
Annex F (informative) Equations for the calculation of Student’s t-values. 122
Annex G (informative) Analysis of variance . 123
Annex H (informative) Equations for the calculation of regression coefficients. 127
Annex I (informative) The intercal method. 129
Bibliography . 130
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 19003 was prepared by Technical Committee ISO/TC 45, Rubber and rubber products, Subcommittee
SC 2, Testing and analysis.
vi © ISO 2006 – All rights reserved
Introduction
Statistical methods have an important role at all stages of the testing process, from the design of the
experiment to the interpretation of results. Hence, those involved in testing require a basic understanding of
statistical principles and knowledge of the statistical techniques which need to be applied.
There are many text books and International Standards which describe statistical methods, but it is convenient
to have a guide which is a single, easy source of reference to the most commonly used methods and formulae,
and which also considers their particular application to the various rubber test methods. This International
Standard is therefore complementary both to the general standards on statistics and to the standards on
methods of test for rubber.
The approach taken in this International Standard is that, for each subject, the text is structured into principles,
methodology and applications to rubber testing. Under principles, the basic concepts of the subject are briefly
outlined. Methodology considers the statistical techniques which can be applied; basic procedures and
formulae are given but, as appropriate, more detailed matter is placed in annexes and, for less commonly
used methods or more advanced treatment, reference is made to other publications. “Applications to rubber
testing” indicates how and where the methods may be applied, and gives examples which are particular to
rubber properties and tests.
INTERNATIONAL STANDARD ISO 19003:2006(E)
Rubber and rubber products — Guidance on the application of
statistics to physical testing
1 Scope
This International Standard provides guidance on the application of statistics to rubber testing. It is intended
not to conflict with or replace existing International Standards covering basic statistical techniques, but rather
to complement them and provide examples of those techniques applied to particular rubber testing situations.
2 References
This International Standard refers to other publications that provide information or guidance. These standards
are listed in the Bibliography.
3 Terms and definitions
For the purposes of this document, the following terms and definitions apply.
NOTE These definitions, which are expressed as far as possible in non-mathematical terms, apply to the main
statistical terminology used. More comprehensive and rigorous lists can be found in the various parts of ISO 3534 and in
the standards dealing with specific statistical techniques indicated in the Bibliography.
3.1
population
totality of data that could (theoretically) be obtained to characterize the property of the rubber, compounding
ingredient or process being measured
3.2
sample
data actually available from the population as a result of an experimental test programme having been
undertaken
3.3
variability
tendency for tests performed on nominally identical test pieces to produce different test results
3.4
arithmetic mean
sum of the (population or sample) data divided by the number of values used
NOTE The “average” is the statistic most frequently used to describe a group of data. There are several kinds of
average and they are often used in common parlance without specifying the type, which can be a source of confusion.
Averages fall into two categories: computational and positional. The arithmetic mean is the most frequently used
computational average. Others are considered in Annex B. Positional averages are the median and mode. The calculation
of the arithmetic mean is given in Equations (1) and (2) in 6.2.2.2.
3.5
median
middle value (or average of the two middle values) when the data in a sample are arranged in numerically
increasing value
3.6
mode
value of the property being measured which occurs with the greatest frequency
3.7
residual
difference (+ or −) between each value and the mean
NOTE The sum of the residuals must be 0.
3.8
variance
arithmetic mean of the squared residuals
3.9
standard deviation
square root of the variance
NOTE The calculation of standard deviation is given in Equations (5) and (6) in 6.2.3.2.1.
3.10
coefficient of variation
ratio of the standard deviation to the mean, generally expressed as a percentage
NOTE The calculation of coefficient of variation is given in Equation (8) in 6.2.3.4.
3.11
range
maximum value minus the minimum value
3.12
standard error
standard deviation of the estimate of the population mean
NOTE The calculation of standard error is given in Equation (7) in 6.2.3.2.3.
3.13
bias
difference between the average statistic and the true value of the parameter it is estimating, arising out of one
or more systematic errors
3.14
accuracy
closeness of agreement between a test result and the accepted reference value
3.15
trueness
closeness of agreement between the average value of a large number of test results and the true or accepted
reference value
NOTE It is usually expressed in terms of bias.
3.16
precision
closeness of agreement between test results
2 © ISO 2006 – All rights reserved
3.17
repeatability
precision obtained under conditions where independent test results are obtained with the same method on
identical test material in the same laboratory by the same operator using the same equipment within a short
interval of time
3.18
reproducibility
precision obtained under conditions where independent test results are produced with the same method on
identical test material in different laboratories with different operators using different equipment
3.19
level of significance
probability of error associated with a significance test
3.20
distribution function
function describing the probability that a random variable will take a value less than or equal to a number x
3.21
density distribution
slope of the distribution function at every value, i.e. the first derivative of the distribution function
3.22
normal distribution
symmetrical “bell-shaped” density distribution which is fully defined by its mean and standard deviation
NOTE It is also known as the Laplace Gauss or Gaussian distribution.
3.23
double exponential distribution
asymmetrical distribution, fully defined by a single “shape” parameter, which has been used to characterize
the distribution of tensile strengths in rubber compounds
3.24
Weibull distribution
symmetrical distribution fully defined by three parameters and found to be useful in characterizing lifetime
tests such as fatigue
3.25
degrees of freedom
number of independent differences between the readings available for an estimate of standard deviation
3.26
confidence interval
range within which a value or parameter can be expected to lie with a given probability
3.27
confidence limits
extreme values of the confidence interval
4 Symbols
a,b,c… Constant coefficients in a regression line
C The coefficient of concordance in Friedman’s test, or Cochran’s quotient when testing variances
for the presence of outliers
C The ith cusum value
i
C Factors used in the derivation of regression coefficients
pq
C The coefficient of variation
v
f(x) A property or parameter which is a function of x or a density distribution function
F The observed value of Snedecor’s F-ratio in a given case
F The statistically critical value for F at a given confidence level and for the given degrees of freedom
cr
for the lesser and greater mean squares
F The F-value for a regression line
r
H /H The null/alternative hypothesis parameter
0 a
K Friedman’s statistic for a rank correlation test
M The mean square for factor z
z
n The number of values in a series
p(x) A probability distribution function
P The plot positions for the graphical presentation of a series of values
m
Q Dixon’s quotient when testing values or means for outliers
r The repeatability of a test method for a particular test or series of tests
(r) The repeatability expressed as a percentage of the mean from a test or series of tests
R The reproducibility of a test method for a particular test or series of tests
(R) The reproducibility expressed as a percentage of the mean from a test or series of tests
s The estimate of the population standard deviation from the available sample
s' the standard deviation of a series of numbers
S The weighted standard error for the combination of two series of values, or the rank sum for a
sample in Friedman’s test
S The total sum of the squares of the differences between individual values and their mean
t
S The sums of squares for factor z
z
t Student’s t-value for a given probability (or confidence level) α
α
U The random uncertainty in a measurement
r
U The systematic uncertainty in a measurement
s
v The number of degrees of freedom for factor z
z
x An individual numerical value, such as the tensile strength of a single test piece
x A single value in a series of values, such as a tensile strength in a set of five replicate values
i
x A single value in a series of values in which two factors are present, such as the tensile strength in
ij
sets of replicates obtained at different temperatures
4 © ISO 2006 – All rights reserved
x The arithmetic mean of a series of numbers, x
i
Z The Z-score in hypothesis testing
α, β The probability of an event occurring
µ The population mean of a distribution
ˆ
µ The estimate of the population mean from the available sample
σ The population standard deviation of a distribution
5 Limitations of test results
5.1 Variability
5.1.1 All measurements are subject to variability. It is necessary to know the sources of variability and make
a reliable estimate of its magnitude. From this information, it should then be possible to judge the reliability of
the results and hence their uncertainty and significance.
5.1.2 The term population is, expressed simply, the total number of objects in a large group (see 3.1). In
testing terms, a population may be, for example, the total number of possible tensile strength results which
could be obtained on a particular rubber compound if every piece of the material made was tested.
5.1.3 A sample is a selected number of, for example, parts or tensile results taken from the population.
NOTE 1 To avoid confusion, sample should not be used to mean test piece.
NOTE 2 Sample can have two meanings:
a) in the physical sense, as in taking five parts from a boxful;
b) in the statistical sense, as in taking five test results.
5.1.4 If five tensile strength measurements are made from a sheet taken from a batch of rubber, an
example of the results which might be obtained is shown in Table 1.
Table 1 — Tensile strength measurements from one batch of rubber
Measurement number Tensile strength
MPa
1 16,8
2 15,4
3 16,3
4 17,7
5 17,6
The sources of variability are:
a) the intrinsic variability of the sheet rubber, arising from the fact that it is not perfectly homogeneous;
b) the variability due to the testing procedure, including test piece preparation, machine accuracy and
operation error.
If several sheets are tested, there is an additional source of variability due to variations in moulding.
If several batches are mixed, two more sources of variation are added:
1) that from the mixing procedure;
2) any variation in compounding ingredients.
If sheets which are nominally the same are given to a number of operators, there is variability due to the
operators.
Similarly, if a number of different test apparatuses are used, variability due to the machines is introduced.
Taking things further, sheets may be tested in different laboratories and between-laboratory variability
introduced.
5.1.5 In practice, the magnitude of variability is minimized by carefully controlling the processing operations
and the test apparatus and procedures. It is never eliminated altogether and inter-laboratory comparisons
have demonstrated that for many rubber tests it can be far greater than was previously thought.
Whatever test is carried out, there is genuine variation due to the material and also variation due to
uncontrolled testing errors. It is often very difficult to separate the two. For example, testing errors can arise
from
a) random variations in test piece geometry due to limitations in cutting precision;
b) variations in the response of the test apparatus;
c) fluctuations in the operator’s performance.
These errors may be large or small and of indeterminate direction so that eventually they tend to cancel out.
More serious is systematic error or bias which is unidirectional, for example the error due to a machine being
wrongly calibrated or an operator consistently misreading a scale.
5.1.6 Testing error apart, the sample of results will not be representative of the whole population if the
physical sample is not representative. Differences between repeat mixes and between repeat mouldings
should be expected because of some variation in the quantities and quality of ingredients used, the efficiency
of mixing and the time of curing, etc. If gross errors are made, some very atypical results are recorded and it is
dangerous to rely heavily on one small sample unless certain that it is representative.
The evaluation of an alternative ingredient by comparison with the standard formula may be considered. The
mixes are uniform, the tester follows the procedures correctly and it is concluded, using statistical methods,
that the new ingredient is an improvement. It is easily forgotten that this conclusion assumes that the samples
of each compound were truly representative of the population. If the variability which would arise from repeat
mixings is rather larger than the testing error, as is often the case, then tests on a series of repeat mixes may
show no difference between the ingredients or even that the new ingredient was worse.
5.2 Accuracy, trueness and precision
Accuracy is the closeness of agreement between a test result and the accepted reference value (see 3.14),
while trueness is the closeness of agreement between the average value of a large number of test results and
the true or accepted reference value (see 3.15). Precision, on the other hand, is the closeness of agreement
between the test results (see 3.16), independent of any reference value that may exist. To keep variability to a
minimum, the test method should be as reproducible as possible, i.e. it should have good precision. However,
having high precision may be of little value if the test has a large bias and hence poor accuracy. Both are
required and indeed they are related in that poor precision (poor reproducibility) will contribute to lowering the
accuracy.
Reproducibility (see 3.18) is the term generally reserved to describe the variation found between different
laboratories, and perhaps also at different times. Repeatability (see 3.17) is used to describe the variation
between repeats in the same laboratory at essentially the same time. It follows that laboratories may exhibit
very good repeatability but, because of bias, the reproducibility between the laboratories is poor.
6 © ISO 2006 – All rights reserved
5.3 Relevance and significance
5.3.1 If accuracy or repeatability were the only interest, testing would be limited to the most accurate or
precise methods. However, the test should be relevant in the sense that the results have a useful meaning in
terms of material or product performance. All tests are not equal: some have more relevance than others in
terms of product performance, material consistency or value as design data. The word significance is
sometimes used to mean relevance and applied to the actual test or property measured, but significance is
used in this International Standard in the statistical sense as in one material being significantly stronger, for
example, than another.
Significance in this sense is concerned with whether observed differences in results are likely to be real or can
reasonably be attributed to chance alone. If the probability of obtaining the observed difference through pure
chance is small, for example less than 1 in 20, then the difference is said to be significant.
5.3.2 The set of tensile strength results quoted in 5.1 could be compared to other sets obtained on different
materials on the same occasion giving, for example, three sets as in Table 2.
Table 2 — Tensile strength measurements from three materials
Measurement number Tensile strength
MPa
Material A Material B Material C
1 16,8 15,6 16,4
2 15,4 16,4 15,4
3 16,3 14,5 14,3
4 17,7 15,8 14,7
5 17,6 16,0 14,4
The averages of the results for materials A and B are higher than that for C but an assessment should be
made as to whether or not they are significantly higher. Without the use of statistical tools it is rather difficult to
make this assessment. In fact, using a test for significance as discussed in 7.2.2 it can be proved that A is
significantly greater than C with 95 % confidence but that A is not significantly different from B, again with
95 % confidence. This is a useful conclusion but its limitations should be appreciated. The statistical tests
prove (with a 1 in 20 chance of being wrong) that results A are significantly greater than C. They do not prove
that material A is stronger than material C. It is known that results from one sheet of one mix may not be
representative of a formulation and these results from a very small test programme should be treated with
caution.
5.3.3 In the above example the differences between the average results were relatively small but tensile
strength can be measured accurately with reasonably small variability so that it is not surprising that 10 %
difference could be proved significant. For other, less reproducible tests a much greater percentage difference
may be needed before the difference can be proved significant. For example, in an electrical resistivity test the
mean value for one material was several times higher than that for a second material but the difference could
not be proved significant. The deduction can be made that significance is not only dependent on the difference
between mean values but also on the amount of variability which is inherent in the test.
6 Distribution of results and measures of central tendency
6.1 Principles
A collection of values, for example individual test results relating to a specific property, are arranged about a
mean value. Usually the distribution of results may be represented by a particular mathematical law such as
the curve shown in Figure 1.
...
NORME ISO
INTERNATIONALE 19003
Première édition
2006-10-01
Caoutchouc et produits à base de
caoutchouc — Lignes directrices pour
l'application des statistiques aux essais
physiques
Rubber and rubber products — Guidance on the application of statistics
to physical testing
Numéro de référence
©
ISO 2006
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.
© ISO 2006
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 2006 – Tous droits réservés
Sommaire Page
Avant-propos. vi
Introduction . vii
1 Domaine d'application. 1
2 Références normatives . 1
3 Termes et définitions. 1
4 Symboles . 4
5 Limites des résultats d’essai. 5
5.1 Variabilité. 5
5.2 Exactitude, justesse et fidélité . 7
5.3 Pertinence et signification . 7
6 Distribution de résultats et mesures de la tendance centrale . 8
6.1 Principes. 8
6.2 Méthodologie. 9
6.2.1 Types de distribution. 9
6.2.2 Mesures de la tendance centrale . 13
6.2.3 Mesures de dispersion. 16
6.2.4 Transformation vers une distribution normale. 18
6.2.5 Essai d’écart par rapport à la normalité . 18
6.3 Applications aux essais du caoutchouc . 21
6.3.1 Généralités . 21
6.3.2 Essai de traction . 21
6.3.3 Fatigue . 23
6.3.4 Conversion en une distribution normale.25
6.3.5 Autres usages de la médiane . 26
7 Limites de confiance et différence significative. 27
7.1 Principes. 27
7.2 Méthodologie. 27
7.2.1 Limites et intervalles de confiance . 27
7.2.2 Différence significative. 32
7.3 Applications aux essais du caoutchouc . 36
7.3.1 Généralités . 36
7.3.2 Limites de confiance et de spécification. 36
7.3.3 Comparaison de résultats. 37
8 Méthodes de rangement . 38
8.1 Principes. 38
8.2 Méthodologie. 38
8.2.1 Test de Friedman . 38
8.2.2 Test basé sur le dénombrement des valeurs extrêmes. 40
8.3 Applications aux essais du caoutchouc . 40
9 Critères de rejet des valeurs aberrantes . 42
9.1 Principes. 42
9.2 Méthodologie. 42
9.2.1 Généralités . 42
9.2.2 Essai de Dixon . 42
9.2.3 Essai de Cochran de la variance. 44
9.3 Applications aux essais du caoutchouc . 47
9.3.1 Généralités . 47
9.3.2 Application de l’essai de Dixon à des résultats individuels. 47
9.3.3 Essai de Cochran de la variance . 48
9.3.4 Essai de Dixon appliqué à un groupe de valeurs moyennes . 49
10 Analyse de variance (ANOVA) . 50
10.1 Principes . 50
10.2 Méthodologie. 50
10.2.1 Généralités. 50
10.2.2 Un facteur avec un nombre égal de répétitions. 50
10.2.3 Un facteur avec un nombre variable de répétitions . 52
10.2.4 Analyse de variance à deux facteurs (et plus). 52
10.3 Applications aux essais du caoutchouc.52
11 Analyse de régression. 56
11.1 Principes . 56
11.2 Méthodologie. 57
11.2.1 Généralités. 57
11.2.2 Moindres carrés linéaires. 57
11.2.3 Moindres carrés quadratiques. 58
11.2.4 Moindres carrés cubiques. 58
11.3 Applications aux essais du caoutchouc.59
11.3.1 Généralités. 59
11.3.2 Effet de la température sur la déformation rémanente après compression . 59
11.3.3 Effet du vieillissement sur la résistance à la traction . 61
11.3.4 Essai température-retrait. 62
12 Incertitude de mesure. 63
12.1 Principes . 63
12.2 Méthodologie. 63
12.2.1 Compilation d’une valeur unique pour l’incertitude. 63
12.2.2 Incertitude aléatoire (U ) . 64
r
12.2.3 Incertitude systématique (U ) . 64
s
12.2.4 Écart d’une valeur unique d’incertitude totale. 66
12.2.5 Établissement du rapport des résultats . 66
12.3 Applications aux essais du caoutchouc.67
13 Échantillonnage . 67
13.1 Principes . 67
13.2 Méthodologie. 68
13.2.1 Généralités. 68
13.2.2 Niveau de qualité acceptable et qualité limite.68
13.2.3 Évaluation de la non-conformité . 68
13.2.4 Niveaux de contrôle . 69
13.2.5 Plans d’échantillonnage par attributs. 70
13.2.6 Échantillonnage aléatoire. 71
13.3 Applications aux essais du caoutchouc.72
14 Nombre d’éprouvettes . 72
14.1 Principes . 72
14.2 Méthodologie. 72
14.3 Applications aux essais du caoutchouc.73
14.3.1 Généralités. 73
14.3.2 Affinage des limites de confiance . 73
14.3.3 Affinage de l’état conforme/non conforme. 74
15 Expression des résultats. 74
15.1 Principes . 74
15.2 Méthodologie. 74
15.2.1 Rapport d’essai . 74
15.2.2 Arrondi . 76
15.3 Applications aux essais du caoutchouc.77
15.3.1 Généralités. 77
15.3.2 Construction d’un histogramme. 77
iv © ISO 2006 – Tous droits réservés
15.3.3 Exemples d’arrondi. 78
16 Déclarations relatives à la fidélité . 79
16.1 Généralités . 79
16.2 Principes. 79
16.3 Méthodologie. 80
16.4 Applications aux essais du caoutchouc . 81
17 Plans d’expériences . 82
17.1 Généralités et principes . 82
17.1.1 Généralités . 82
17.1.2 Principes. 84
17.2 Méthodologie. 97
17.2.1 Généralités . 97
17.2.2 Expériences descriptives. 98
17.2.3 Expériences comparatives . 99
17.2.4 Expériences de réponse . 101
17.3 Applications aux essais du caoutchouc . 101
17.3.1 Expériences descriptives. 101
17.3.2 Expériences comparatives . 103
17.3.3 Expériences de réponse . 107
18 Maîtrise statistique de la qualité . 112
18.1 Principes. 112
18.2 Méthodologie. 112
18.2.1 Généralités . 112
18.2.2 Cartes de contrôle utilisant des attributs. 112
18.2.3 Cartes de contrôle utilisant des mesures . 112
18.3 Applications aux essais du caoutchouc . 114
18.3.1 Généralités . 114
18.3.2 Cartes de contrôle . 114
18.3.3 Carte à somme cumulée . 117
Annexe A (informative) Forme mathématique des fonctions de distribution référencées dans la
présente Norme internationale. 121
Annexe B (informative) Autres formes de la valeur moyenne. 123
Annexe C (informative) Interrelations relatives aux mesures de la tendance centrale dans la loi
doublement exponentielle et la loi de Weibull. 124
Annexe D (informative) Équation pour le calcul de l'écart-type . 125
Annexe E (informative) Construction du papier à échelle fonctionnelle de Weibull. 127
Annexe F (informative) Équations pour le calcul des valeurs t de Student . 128
Annexe G (informative) Analyse de la variance . 129
Annexe H (informative) Équations pour le calcul des coefficients de régression . 133
Annexe I (informative) Méthode intercal . 135
Bibliographie . 136
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 19003 a été élaborée par le comité technique ISO/TC 45, Élastomères et produits à base d'élastomères,
sous-comité SC 2, Essais et analyses.
vi © ISO 2006 – Tous droits réservés
Introduction
Les méthodes statistiques jouent un rôle majeur au cours des différentes phases du processus d’essai, de la
conception de l’expérience à l’interprétation des résultats. Par conséquent, tous les acteurs intervenant dans
les essais doivent disposer de connaissances élémentaires des principes statistiques et appréhender les
techniques statistiques à mettre en œuvre.
Il existe de nombreux ouvrages et Normes internationales qui décrivent des méthodes statistiques mais il
semble judicieux de pouvoir disposer d’un guide qui puisse être utilisé comme source de référence unique et
conviviale pour les méthodes et les formules les plus communément appliquées, et qui s’attache également à
leur application particulière aux diverses méthodes d’essai du caoutchouc. En conséquence, la présente
Norme internationale complète à la fois les normes génériques sur les statistiques et les normes traitant des
méthodes d’essai applicables aux caoutchoucs.
L’approche adoptée dans la présente Norme internationale est telle que, pour chaque thème abordé, le texte
est subdivisé en principes, méthodologie et applications aux essais du caoutchouc. Les concepts de base du
thème en question sont brièvement décrits sous l’intitulé «Principes». Le volet «Méthodologie» s’attache aux
techniques statistiques qui peuvent être appliquées et présente quelques procédures de base et formules, les
questions plus spécifiques étant placées en annexe, et pour les méthodes moins couramment utilisées ou les
traitements plus avancés, il est fait référence à d’autres publications. Le volet «Applications aux essais du
caoutchouc» précise le mode et les modalités d’application de ces méthodes, et donne des exemples propres
aux propriétés et aux essais du caoutchouc.
NORME INTERNATIONALE ISO 19003:2006(F)
Caoutchouc et produits à base de caoutchouc — Lignes
directrices pour l'application des statistiques aux essais
physiques
1 Domaine d'application
La présente Norme internationale fournit des lignes directrices sur l’application des statistiques aux essais du
caoutchouc. Elle est destinée à éviter tout conflit avec des Normes internationales existantes traitant de
techniques statistiques de base et ne vise en aucun cas à se substituer à ces normes, son unique objectif
étant de compléter ces dernières et d’illustrer par des exemples les techniques appliquées à des situations
d’essai particulières du caoutchouc.
2 Références normatives
La présente Norme internationale fait référence à d’autres publications qui fournissent des informations ou
des lignes directrices. Ces normes sont répertoriées dans la Bibliographie.
3 Termes et définitions
Pour les besoins du présent document, les termes et définitions suivants s'appliquent.
NOTE Ces définitions, qui sont autant que possible exprimées en termes non mathématiques, s’appliquent à la
terminologie statistique essentielle utilisée. Des listes plus complètes et plus rigoureuses peuvent être trouvées dans les
diverses parties de l’ISO 3534 et dans les normes traitant de techniques statistiques spécifiques mentionnées dans la
Bibliographie.
3.1
population
totalité des données qui pourraient (en théorie) être obtenues pour caractériser la propriété du caoutchouc, de
l’ingrédient de mélange ou du processus mesuré
3.2
échantillon
données effectivement disponibles au départ de la population résultant d’un programme d’essai expérimental
mis en œuvre
3.3
variabilité
tendance qu’ont des éprouvettes nominalement identiques à produire des résultats d’essai différents
3.4
moyenne arithmétique
somme des données (de la population ou de l’échantillon) divisée par le nombre de valeurs utilisées
NOTE La «moyenne» est la statistique la plus fréquemment utilisée pour décrire un groupe de données. Il existe
plusieurs types de moyenne qui sont fréquemment utilisés en langage courant sans en préciser le type, ce qui peut être
source de confusion. Les moyennes sont classées en deux catégories: celles définies par un calcul et celles définies par
un rang. La moyenne arithmétique est la moyenne définie par un calcul la plus fréquemment utilisée. D’autres moyennes
sont examinées dans l’Annexe B. Les moyennes définies par un rang sont la médiane et le mode. Le calcul de la
moyenne arithmétique est donné dans les Équations (1) et (2) au 6.2.2.2.
3.5
médiane
valeur centrale (ou moyenne des deux valeurs centrales) lorsque les données d’un échantillon sont classées
par ordre numérique croissant
3.6
mode
valeur de la propriété mesurée qui survient avec la fréquence maximale
3.7
résidu
différence (+ ou −) entre chaque valeur et la moyenne
NOTE La somme des résidus doit être égale à 0.
3.8
variance
moyenne arithmétique du carré des résidus
3.9
écart-type
racine carrée de la variance
NOTE Le calcul de l’écart-type est donné dans les Équations (5) et (6) au 6.2.3.2.1.
3.10
coefficient de variation
rapport de l’écart-type à la moyenne, généralement exprimé en pourcentage
NOTE Le calcul du coefficient de variation est donné dans l’Équation (8) au 6.2.3.4.
3.11
étendue
valeur maximale moins la valeur minimale
3.12
erreur type
écart-type de l’estimation de la moyenne d’une population
NOTE Le calcul de l’erreur type est donné dans l’Équation (7) au 6.2.3.2.3.
3.13
biais
différence entre la statistique moyenne et la valeur vraie du paramètre qu’elle estime, résultant d’une ou
plusieurs erreurs systématiques
3.14
exactitude
étroitesse de l’accord entre le résultat d’essai et la valeur de référence acceptée
3.15
justesse
étroitesse de l’accord entre la valeur moyenne d’un grand nombre de résultats d’essai et la valeur de
référence acceptée ou vraie
NOTE Elle est généralement exprimée en termes de biais.
2 © ISO 2006 – Tous droits réservés
3.16
fidélité
étroitesse d’accord entre des résultats d’essai
3.17
répétabilité
fidélité obtenue sous des conditions où les résultats d’essai indépendants sont obtenus par la même méthode
sur un matériau d’essai identique dans le même laboratoire, par le même opérateur, utilisant le même
équipement et pendant un court intervalle de temps
3.18
reproductibilité
fidélité obtenue sous des conditions où les résultats d’essai indépendants sont obtenus par la même méthode
sur un matériau d’essai identique dans différents laboratoires, avec différents opérateurs et utilisant des
équipements différents
3.19
niveau de signification
probabilité d’erreur associée à un test de niveau de signification
3.20
fonction de répartition
fonction décrivant la probabilité qu’une variable aléatoire prenne une valeur inférieure ou égale à un nombre x
3.21
distribution en densité
pente de la fonction de répartition à chaque valeur, c’est-à-dire la dérivée première de la fonction de
répartition
3.22
distribution normale
distribution en densité de probabilité «en cloche» symétrique qui est totalement définie par sa moyenne et son
écart-type
NOTE Elle est également connue comme distribution de Laplace-Gauss ou distribution gaussienne.
3.23
loi doublement exponentielle
distribution asymétrique, totalement définie par un unique paramètre «forme», qui a été utilisé pour
caractériser la distribution des résistances à la traction dans les mélanges de caoutchouc
3.24
loi de Weibull
distribution symétrique totalement définie par trois paramètres et généralement utilisée pour caractériser les
essais de durée de vie, tels que les essais de fatigue
3.25
degrés de liberté
nombre de différences indépendantes entre les lectures disponibles pour une estimation de l’écart-type
3.26
intervalle de confiance
étendue dans laquelle une valeur ou un paramètre peut être supposé(e) se trouver avec une probabilité
donnée
3.27
limites de confiance
valeurs extrêmes de l’intervalle de confiance
4 Symboles
a,b,c… Coefficients constants d’une droite de régression
C Coefficient de concordance dans le test de Friedman, ou quotient de Cochran lors du test des
variances visant à établir la présence de valeurs aberrantes
e
C i valeur à somme cumulée
i
C Facteurs utilisés dans la dérivée des coefficients de régression
pq
C Coefficient de variation
v
f(x) Propriété ou paramètre qui est une fonction de x ou une distribution en densité
F Valeur observée du rapport F de Snedecor dans une situation donnée
F Valeur statistiquement critique pour «F» à un niveau de confiance donné et pour les degrés de
cr
liberté donnés pour les valeurs quadratiques moyennes inférieures et supérieures
F Valeur F pour une droite de régression
r
H /H Paramètre d’une hypothèse nulle/alternative
0 a
K Statistique de Friedman pour un test de corrélation de rang
M Valeur quadratique moyenne pour le facteur z
z
n Nombre de valeurs dans une série
p(x) Fonction de répartition de probabilité
P Positions graphiques pour la représentation graphique d’une série de valeurs
m
Q Quotient de Dixon lors du test de valeurs ou de moyennes pour les valeurs aberrantes
r Répétabilité d’une méthode d’essai pour un test particulier ou une série de tests
(r) Répétabilité exprimée en pourcentage de la moyenne à partir d’un test ou d’une série de tests
R Reproductibilité d’une méthode d’essai pour un test particulier ou une série de tests
(R) Reproductibilité exprimée en pourcentage de la moyenne à partir d’un test ou d’une série de tests
s Estimation de l’écart-type de population à partir de l’échantillon disponible
′
s Écart-type d’une série de nombres
S Erreur type pondérée pour la combinaison de deux séries de valeurs, ou somme des rangs pour un
échantillon dans le test de Friedman
S Somme totale des carrés des différences entre des valeurs individuelles et leur moyenne
t
S Somme des carrés pour le facteur z
z
t Valeur t de Student pour une probabilité donnée (ou niveau de confiance) α
α
U Incertitude aléatoire d’une mesure
r
U Incertitude systématique d’une mesure
s
v Degrés de liberté pour le facteur z
z
x Valeur numérique individuelle, telle que la résistance à la traction d’une seule éprouvette
x Valeur unique d’une série de valeurs, telle que la résistance à la traction dans un ensemble de
i
cinq répétitions
4 © ISO 2006 – Tous droits réservés
x Valeur unique d’une série de valeurs dans laquelle deux facteurs sont présents, tels que la
ij
résistance à la traction dans des ensembles de répétitions obtenues à différentes températures
x Moyenne arithmétique d’une série de nombres, x
i
Z Note Z du test d’hypothèse
α, β Probabilité d’occurrence d’un événement
µ Moyenne de population d’une distribution
ˆ
µ Estimation de la moyenne de population à partir de l’échantillon disponible
σ Écart-type de population d’une distribution
5 Limites des résultats d’essai
5.1 Variabilité
5.1.1 Toutes les mesures sont sujettes à variabilité. Il est donc nécessaire de connaître les sources de
variabilité et d’établir une estimation fiable de son ampleur. À partir de ces informations, il devrait alors être
possible d’apprécier la fiabilité des résultats obtenus et, par conséquent, leur incertitude et leur signification.
5.1.2 La notion de population, exprimée simplement, est le nombre total d’objets dans un groupe important
(voir 3.1). En termes d’essais, une population peut, par exemple, correspondre au nombre total de résultats
possibles de résistance à la traction qui pourraient être obtenus sur un mélange de caoutchouc particulier si
chaque élément du matériau était soumis à essai.
5.1.3 Un échantillon est un nombre sélectionné, par exemple, d’individus ou de résultats d’essai en traction
prélevés à partir de la population.
NOTE 1 Pour éviter toute confusion, il convient de ne pas utiliser le terme «échantillon» pour faire référence à une
éprouvette.
NOTE 2 «Échantillon» peut avoir deux sens:
a) au sens physique, prélèvement de cinq individus dans une boîte;
b) au sens statistique, prélèvement de cinq résultats d’essai.
5.1.4 Le Tableau 1 présente un exemple des résultats susceptibles d’être obtenus, dans l’hypothèse où il
est procédé à cinq mesurages de la résistance à la traction à partir d’une feuille prélevée au départ d’un lot de
caoutchouc.
Tableau 1 — Mesurages de la résistance à la traction réalisés à partir d’un lot de caoutchouc
Numéro de mesurage Résistance à la traction
MPa
1 16,8
2 15,4
3 16,3
4 17,7
5 17,6
Les sources de variabilité sont:
a) la variabilité intrinsèque du caoutchouc en feuilles imputable au fait que celui-ci n’est pas parfaitement
homogène;
b) la variabilité imputable au mode opératoire d’essai, y compris la préparation des éprouvettes, l’exactitude
de la machine et l’erreur opératoire.
Si plusieurs feuilles sont soumises à essai, les variations de moulage constituent une source supplémentaire
de variabilité.
Si plusieurs lots sont mélangés, deux sources complémentaires de variation viennent s’ajouter:
1) la variation résultant de l’opération de mélangeage;
2) toute variation des ingrédients de mélange.
Si des feuilles, nominalement identiques, sont confiées à plusieurs opérateurs, il y a variabilité en raison de
ces opérateurs.
De même, s’il est fait usage de plusieurs appareillages d’essai différents, ces machines sont autant de
sources de variabilité. En poursuivant le raisonnement, il est possible que les feuilles soient mises à essai
dans différents laboratoires, auquel cas il y a introduction d’une variabilité interlaboratoires.
5.1.5 En pratique, l’ampleur de la variabilité est réduite par une maîtrise minutieuse des opérations de
traitement ainsi que de l’appareillage et des modes opératoires d’essai. La variabilité ne peut cependant
jamais être totalement éliminée; à cet égard, des comparaisons interlaboratoires ont révélé que, pour un
grand nombre d’essais de caoutchouc, elle peut être nettement supérieure aux valeurs autrefois
communément supposées.
Quel que soit l’essai réalisé, une véritable variation due au matériau et également la variation imputable aux
erreurs d’essai non maîtrisées ne peuvent être évitées. Il est souvent très difficile de distinguer ces deux
variations. Par exemple, les erreurs d’essai peuvent être le résultat de:
a) variations aléatoires de la géométrie des éprouvettes en raison des limites de précision de coupe;
b) variations de la réponse de l’appareillage d’essai;
c) fluctuations des performances de l’opérateur.
Ces erreurs peuvent être importantes ou minimes et de sens indéterminé, de sorte qu’elles tendent finalement
à s’annuler. Plus sérieux est l’erreur systématique ou le biais qui est unidirectionnel, par exemple l’erreur due
à un étalonnage incorrect d’une machine ou le fait qu’un opérateur lise invariablement une échelle de façon
erronée.
6 © ISO 2006 – Tous droits réservés
5.1.6 L’erreur d’essai mise à part, l’échantillon de résultats ne sera pas représentatif de l’ensemble de la
population si l’échantillon physique n’est pas représentatif. Il y a lieu de s’attendre à des différences entre des
mélanges répétés et entre des moulages répétés en raison d’une certaine variation des quantités et de la
qualité des ingrédients utilisés, de l’efficacité du mélangeage, du temps de durcissement, etc. Si des erreurs
grossières sont commises, quelques résultats très atypiques sont enregistrés et il est dangereux de se fier
excessivement à un seul petit échantillon à moins d’être assuré que celui-ci est représentatif.
L’évaluation d’un ingrédient de substitution par comparaison avec la formulation normalisée peut être
envisagée. Les mélanges sont uniformes, le formulateur suit correctement les procédures et il est conclu, au
moyen de méthodes statistiques, que le nouvel ingrédient est une amélioration. Il est aisément omis que cette
conclusion suppose que les échantillons de chaque mélange étaient véritablement représentatifs de la
population. Si la variabilité résultant de mélanges répétés est relativement plus importante que l’erreur d’essai,
comme cela est souvent le cas, il est possible que les essais réalisés sur une série de mélangeages répétés
ne montrent aucune différence entre les ingrédients ou qu’ils indiquent que le nouvel ingrédient est un retour
en arrière.
5.2 Exactitude, justesse et fidélité
L’exactitude est l’étroitesse de l’accord entre un résultat d’essai et la valeur de référence acceptée (voir 3.14),
tandis que la justesse correspond à l’étroitesse de l’accord entre la valeur moyenne obtenue à partir d’une
grande série de résultats et la valeur de référence acceptée ou vraie (voir 3.15). En revanche, la fidélité est
l’étroitesse de l’accord entre les résultats d’essai (voir 3.16), indépendamment de toute valeur de référence
susceptible d’exister. Pour maintenir la variabilité à un minimum, il convient que la méthode d’essai soit aussi
reproductible que possible, c’est-à-dire présente une bonne fidélité. Toutefois, cette fidélité élevée ne
présente qu’un faible intérêt si l’essai est caractérisé par un biais important et, de ce fait, par une exactitude
médiocre. Ces deux éléments sont obligatoires et étroitement corrélés en ce sens qu’une faible fidélité (faible
reproductibilité) contribuera à une réduction de l’exactitude.
La reproductibilité (voir 3.18) est le terme généralement réservée à la description de la variation observée
entre différents laboratoires et, éventuellement, aussi à différents moments. La répétabilité (voir 3.17) est
quant à elle utilisée pour décrire la variation entre les répétitions dans un même laboratoire à un moment
pratiquement identique. Il s’ensuit que des laboratoires peuvent faire preuve d’une très bonne répétabilité,
mais que la reproductibilité entre ces laboratoires est médiocre en raison du biais.
5.3 Pertinence et signification
5.3.1 Si l’exactitude ou la répétabilité étaient les seuls éléments présentant de l’intérêt, les essais se
limiteraient aux méthodes les plus exactes ou fidèles. Il convient cependant que l’essai soit pertinent au sens
où les résultats ont une signification utile en termes de performances du matériau ou du produit. Tous les
essais ne sont pas égaux, certains ont une pertinence supérieure à d’autres en termes de performances du
produit, de résistance du matériau ou de valeur caractéristiques utile à la conception. Le vocable
«signification» est parfois utilisé pour faire référence à la pertinence et appliqué à l’essai réel ou à la propriété
mesurée; toutefois, la signification évoquée dans la présente Norme internationale est prise au sens
statistique comme dans l’exemple suivant: un matériau est significativement plus résist
...














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...