ISO 3951:1989
(Main)Sampling procedures and charts for inspection by variables for percent nonconforming
Sampling procedures and charts for inspection by variables for percent nonconforming
Is complementary to ISO 2859. Object of the methods laid down is to ensure that lots of an acceptable quality have a high probability of acceptance and that the probability of rejection of inferior lots is as high as possible. To be used under the following conditions: inspection procedure to be applied to a continuous series of lots of discrete products supplied by one producer using one production process; only a single quality characteristic is considered which must be measurable on a continuous scale; production is stable and quality characteristic is distributed normally; a contract or standard defines an upper and/or a lower specification limit.
Règles et tables d'échantillonnage pour les contrôles par mesures des pourcentages de non conformes
Sampling procedures and charts for inspection by variables for percent nonconforming
General Information
- Status
- Withdrawn
- Publication Date
- 13-Sep-1989
- Withdrawal Date
- 13-Sep-1989
- Technical Committee
- ISO/TC 69/SC 5 - Acceptance sampling
- Current Stage
- 9599 - Withdrawal of International Standard
- Start Date
- 31-Mar-2006
- Completion Date
- 12-Feb-2026
Relations
- Effective Date
- 11-Feb-2026
- Effective Date
- 11-Feb-2026
- Effective Date
- 11-Feb-2026
- Effective Date
- 10-Feb-2026
- Effective Date
- 10-Feb-2026
- Effective Date
- 10-Feb-2026
- Effective Date
- 10-Feb-2026
- Effective Date
- 10-Feb-2026
- Effective Date
- 10-Feb-2026
- Effective Date
- 09-Feb-2026
- Effective Date
- 09-Feb-2026
- Referred By
EN 12654-2:1998 - Textile glass - Yarns - Part 2: Methods of test and general specifications - Effective Date
- 09-Feb-2026
- Effective Date
- 09-Feb-2026
- Effective Date
- 09-Feb-2026
- Effective Date
- 09-Feb-2026
ISO 3951:1989 - Sampling procedures and charts for inspection by variables for percent nonconforming
ISO 3951:1989 - Regles et tables d'échantillonnage pour les contrôles par mesures des pourcentages de non conformes
ISO 3951:1989 - Regles et tables d'échantillonnage pour les contrôles par mesures des pourcentages de non conformes
Get Certified
Connect with accredited certification bodies for this standard

BSI Group
BSI (British Standards Institution) is the business standards company that helps organizations make excellence a habit.

Bureau Veritas
Bureau Veritas is a world leader in laboratory testing, inspection and certification services.

DNV
DNV is an independent assurance and risk management provider.
Sponsored listings
Frequently Asked Questions
ISO 3951:1989 is a standard published by the International Organization for Standardization (ISO). Its full title is "Sampling procedures and charts for inspection by variables for percent nonconforming". This standard covers: Is complementary to ISO 2859. Object of the methods laid down is to ensure that lots of an acceptable quality have a high probability of acceptance and that the probability of rejection of inferior lots is as high as possible. To be used under the following conditions: inspection procedure to be applied to a continuous series of lots of discrete products supplied by one producer using one production process; only a single quality characteristic is considered which must be measurable on a continuous scale; production is stable and quality characteristic is distributed normally; a contract or standard defines an upper and/or a lower specification limit.
Is complementary to ISO 2859. Object of the methods laid down is to ensure that lots of an acceptable quality have a high probability of acceptance and that the probability of rejection of inferior lots is as high as possible. To be used under the following conditions: inspection procedure to be applied to a continuous series of lots of discrete products supplied by one producer using one production process; only a single quality characteristic is considered which must be measurable on a continuous scale; production is stable and quality characteristic is distributed normally; a contract or standard defines an upper and/or a lower specification limit.
ISO 3951:1989 is classified under the following ICS (International Classification for Standards) categories: 03.120.30 - Application of statistical methods. The ICS classification helps identify the subject area and facilitates finding related standards.
ISO 3951:1989 has the following relationships with other standards: It is inter standard links to EN ISO/IEC 15415:2005, CEN ISO/TS 15875-7:2003, CEN ISO/TS 15874-7:2003, EN 61284:1997, EN 60432-1:2000, EN 61897:1998, EN 61854:1998, EN 60086-3:2005, EN 60086-1:2001, EN 206-1:2000, EN 681-3:2000, EN 12654-2:1998, EN 12763:2000, EN 13084-5:2005, EN 13975:2003. Understanding these relationships helps ensure you are using the most current and applicable version of the standard.
ISO 3951:1989 is available in PDF format for immediate download after purchase. The document can be added to your cart and obtained through the secure checkout process. Digital delivery ensures instant access to the complete standard document.
Standards Content (Sample)
INTERNATIONAL Is0
STANDARD 3951
Second edition
1989-09-15
Sampling- procedures and charts for inspection
by variables for percent nonconforming
R&g/es et tables d ’khan tiilonnage pour /es contr&es par mesures des pourcentages
de non con formes
Reference number
IS0 3951 : 1989 (El
IS0 3951 : 1989 (El
Foreword
IS0 (the International Organization for Standardization) is a worldwide federation of
national standards bodies (IS0 member bodies). The work of preparing International
Standards is normally carried out through IS0 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, govern-
mental and non-governmental, in liaison with ISO, also take part in the work. IS0
collaborates closely with the International Electrotechnical Commission (IEC) on all
matters of electrotechnical standardization.
Draft International Standards adopted by the technical committees are circulated to
the member bodies for approval before their acceptance as International Standards by
the IS0 Council. They are approved in accordance with IS0 procedures requiring at
least 75 % approval by the member bodies voting.
International Standard IS0 3951 was prepared by Technical Committee ISO/TC 69,
Applications of stat/s tical methods.
This second edition cancels and replaces the first edition (IS0 3951 : 1981), of which it
constitutes a technical revision.
The principal changes to the first edition are as follows:
a) a distinction has been drawn between maximum process standard deviation
(MPSD) under the “a” method and maximum sample standard deviation (MSSD)
under the l ’s” method;
b) all the “a” method acceptance curves have been truncated at the appropriate
maximum process standard deviations;
c) the terminology has been aligned on that of IS0 2859 and IS0 3534.
Users should note that all International Standards undergo revision from time to time
and that any reference made herein to any other International Standard implies its
latest edition, unless otherwise stated.
0 IS0 1989
All rights reserved. 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 the publisher.
International Organization for Standardization
Case postale 56 l CH-1211 Geneve 20 l Switzerland
Printed in Switzerland
ii
IS0 3951 : 1989 (E)
Contents
Page
Section one : General
.................................................
1 Scope and field of application
........................................
........................................................
2 References
3 Definitions and symbols.
............................................
4 Acceptable Quality Level (AQL) . ; .
5 Switching rules for normal, tightened and reduced inspection. .
6 Relation to IS0 2959.
...............................................
7 Non-continuous production and operating characteristic curves. .
Section two : Choice of sampling plan
.................................
8 Planning .
9 Choice between variables and attributes
...............................
10 Choice of method
..................................................
II Choice of inspection level and AQL
...................................
12 Choice of sampling plan
.............................................
Section three : Operation of a variables sampling plan
.................. 8
13 Preliminary operations 8
..............................................
14 Standard procedure for “s” method.
.................................. 8
15 Standard procedure for “0” method
.................................. 12
Procedure during continuing inspection
16 . 15
17 Normality and outliers
.............................................. 15
Records .
18 15
19 Operation of switching rules
......................................... 16
Discontination and resumption of inspection
20 . 16
21 Acceptance curves for tightened and reduced inspection.
................ 16
Switching to “a” method
22 . 16
. . .
III
IS0 3951 : 1989 (E)
Page
Section four : Tables and diagrams . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
Tables
I-A Sample size code letters and inspection levels . . . . . . . . . . . . . . 18
I-B Sample size code letters and sample sizes for normal
inspection. 19
Single sampling plans for normal inspection
II-A
(master table) : l ’s” method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
II-B Single sampling plans for tightened inspection
(master table) : “s” method . . . . . . . . i . . . . . . . . . 21
Single sampling plans for reduced inspection
II-C
(master table) : l ’s” method. . . . . . . . . . . . . . . . . . 22
Ill-A Single sampling plans for normal inspection
(master table) : “a” method . . . . . . . . . . . . . . . . . 23
Single sampling plans for tightened inspection
III-B
(master table) : “a” method . . . . . . . . . . . . . . . . . 24
III-C Single sampling plans for reduced inspection
(master table) : “a” method . . . . . . . . . . . . . . . . . . . 25
Values of “fS,, for maximum sample standard
IV-S
deviation (MSSD) : “s ”method . . . . . . . . . . . . . . 26
IV-0 Values of “f0,’ for maximum process standard
deviation (MPSD) : “a” method . . . . . . . . . . . . . . . . . . . . . . . . 26
V (V-B to V-P) Tables and operating characteristic curves for single
sampling plans (sample size code letters B to P) 27-40
Correspondence between tig htened and normal
VI-A
inspection plans. 41
VI-B Correspondence between reduced and normal
inspection plans.
Diagrams
Sample size code letters of standard sampling plans for specified
A
qualities at 95 % and 10 % probabilities of acceptance . . . . . . .
s-D to s-P Acceptance curves for combined double specification
limits : ‘Is” method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43-54
Acceptance curves for combined double specification
a-c to a-P
limits : “a” method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55-67
Annexes
.....................................
A Procedures for obtaining s and 0 68
..................................................
B Statistical theory. 71
C Sampling plans for “R” method . 82
......................................
D Graph paper for the I ’s” method 105
Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107
IV
INTERNATIONAL STANDARD
IS03951 : 1989 (E)
Sampling procedures and charts for inspection by
variables for percent nonconforming
Section one : General
normal distribution or a close approximation to the normal
1 Scope and field of application
distribution;
d) where a contract or standard defines an upper
1.1 Scope
specification limit U, a lower specification limit L, or
both; a product is qualified as nonconforming when its
1 .I .I This International Standard establishes sampling plans
measured quality characteristic x satisfies one of the fol-
and procedures for inspection by variables. It is complementary
lowing inequalities :
to IS0 2859. When specified by the responsible authority, both
x>u . . . (1)
this International Standard and IS0 2859 may be referenced in
a product or process specification contract, inspection instruc-
x
tions’ or other documents and the provisions set forth therein
eitherx > Uorx < L
. . . (3)
shall govern. The “responsible authority” shall be designated in
one of the above documents.
Inequalities (1) and (2) are called cases with a single specifica-
tion limit, and (3) a case with double specification limits. In
this last situation a further distinction is made between separate
1.1.2 The object of the methods laid down in this Inter-
or combined double limits according to whether the AQL is
national Standard is to ensure that lots of an acceptable quality
applied to each limit separately or to both limits combined (see
have a high probability of acceptance and that the probability
clause 4).
of not accepting inferior lots is as high as possible.
1.1.3 In common with IS0 2859, the percentage of noncon-
2 References
forming products in the lots is used to define the quality of
these lots and of the production process in question. I SO 2854, Statistical interpretation of data - Techniques of
estimation and tests relating to means and variances.
1.2 Field of application IS0 2859, Sampling procedures and tables for inspection by
attributes.
This International Standard is primarily designed for use under
IS0 3534, Statistics - Vocabulary and symbols.
the following conditions :
I SO 5479, Normality tests. 1)
a) where the inspection procedure is to be applied to a
IS0 5725, Precision of test methods - Determination of
continuous series of lots of discrete products all supplied
repeatability and reproducibility for a standard test method by
by one producer using one production process. If there are
inter-laboratory tests.
different producers this International Standard shall be ap-
plied to each one separately;
b) where only a single quality characteristic x of these
3 Definitions and symbols
products is taken into consideration, which must be
measurable on a continuous scale. If several such
3.1 Definitions
characteristics are of importance, this International Stan-
dard shall be applied to each separately;
For the purposes of this International Standard, the definitions
c) where production is stable (under statistical control) given in IS0 3534 and IS0 2859, together with the following,
and the quality characteristic x is distributed according to a
apply.
1) At present at the stage of draft.
IS0 3951 :I989 (E)
3.1 .I0 specification limit (limite de specification) : The
3.1.1 inspection by variables (controle par mesures) : A
limiting value (lower or upper) specified for a quantitative
method which consists in measuring a quantitative character-
characteristic.
istic for each item of a population or a sample taken from
this population.
3.1 .I1 lower specification limit (L) [limite inferieure de spe-
cification (Li)]: A specification limit that defines the lower con-
3.1.2 acceptance sampling by variables (echantillonnage
formance boundary for an individual unit of a manufacturing or
par mesures en vue d ’acceptation) : An acceptance procedure
service operation.
wherein a specified characteristic is measured to establish
statistically the acceptability of a lot from the result obtained
3.1.12 upper specification limit (U) [limite superieure de
from the items in a sample.
specification (L,)]: A specification limit that defines the upper
conformance boundary for an individual unit of a manufactur-
3.1.3 acceptable quality level (AQL) [niveau de qualite ac-
ing or service operation.
ceptable (NQA)] : When a continuous series of lots is con-
sidered, a quality level which for the purposes of sampling in-
n limit (limite unique de specifica-
3.1.13 single specificatio
spection is the limit of a satisfactory process average percent
tion) : The term used when one limit only is specified.
nonconforming. (See clause 4.)
3.1.14 separate double specification limits (limites de
3.1.4 limiting quality (qualite limite) : When a lot is con-
specifications doubles separees) : The term used when both
sidered in isolation, a quality level which for the purposes of
upper and lower limits are specified and separate AQLs are
sampling inspection is limited to a low probability of accep-
applied to each limit individually. (See 4.3.)
tance (in this International Standard : 10 %). (See 12.1.)
3.1.15 combined double specification limit (limite de
specification double combinee) : The term used when both
3.1.5 nonconformity (non-conformite) : Failure to fulfil a
specified requirement by a quality characteristic of an item or upper and lower limits are specified and an AQL is given which
service, the assessment of which does not depend essentially applies to the combined percent nonconforming at the two
on the passage of time. limits. (See 4.3.)
Nonconformities will generally be classif ied by their degree of
3.1 .I6 acceptability constant (k) [constante d ’acceptabilite
seriousness, such as
(HI : A constant dependent on the specified value of the ac-
ceptable quality level and the sample size. (See 14.2 and 15.2,
or clause C.5 in annex C.)
Class A. Those nonconformities of a type considered to be of
the highest concern for the product or service. In acceptance
sampling, such types of nonconformity will be assigned very 3.1.17 quality statistic (Q) [statistique de qualite (Q)] : A
small AQL values. function of the specification limit, the sample mean, and the
standard deviation. The lot is sentenced on the result of com-
paring Q with the acceptability constant k. (See 14.2 and 15.2,
Class B. Those nonconformities of a type considered to have
or clause C.5 in annex C.)
the next lower degree of concern; therefore these can be
assigned a larger AQL value than those in class A and smaller
3.1.18 lower quality statistic (0,) [statistique de qualite
than in class C, if a third class exists, and so on.
correspondant a la limite inferieure (Qi)] : A function of the
lower specification limit, the sample mean, and the standard
The number of classes and the assignment into a class should
deviation. The lot is sentenced on the result of comparing Q,
be appropriate to the quality requirements of the specific situa-
with the acceptability constant k. (See 14.2 and 15.2, or clause
tion.
C.5 in annex C.)
3.1.6 nonconforming u nit (unite non conforme): A unit of
3.1.19 upper quality statistic (QJ [statistique de qualite
product or service contain ing at least one nonconformity.
correspondant a la limite superieure (QJ] : A function of the
upper specification limit, the sample mean, and the standard
3.1.7 “s” method (methode ((SD) : A method of assessing the
deviation. The lot is sentenced on the result of comparing QU
acceptability of a lot by using the sample standard deviation.
with the acceptability constant k. (See 14.2 and 15.2, or clause
(See clause 14.1
C.5 in annex C.)
3.1.8 “o“ method (methode ((0))) : A method of assessing
3.1.20 maximum sample standard deviation (MSSD)
the acceptability of a lot by using knowledge of the process
[&art-type maximal de I ’echantillon (ETME)]: Under given con-
standard deviation, (See clause 15.)
ditions, the largest acceptable sample standard deviation. (See
14.6 and B.8.3 in annex B.)
3.1.9 “R” method (methode
the acceptability of a lot by indirectly using an estimate of the
3.1.21 maximum process standard deviation (MPSD)
standard deviation of the process based on the average range [&art-type maximal du procede (ETMP)]: Under given con-
of the measurements of the items in sub-groups of the sample.
ditions, the largest acceptable process standard deviation. (See
(See annex C.) 15.3 and B.5.2 in annex B).
IS0 3951 : 1989 (E)
> “Greater than” (for example, a > b means a is greater
3.122 switching rules (regles de modification du controle) :
than 6).
Instructions within a sampling scheme for shifting from one
sampling plan to another based on demonstrated quality
“Greater than or equal to” (for example, a > b means a
history. (See clause 19.) >
is greater than or equal to b).
3.2 Symbols
< “Less than” (for example, a < b means a is less than
b).
The symbols used are as follows:
“Less than or equal to” (for example, a < b
G a is
, ‘s A factor, given in table IV-s, that relates the maximum
less than or equal to 6).
sample standard deviation to the difference between U
and L.
3.3 Bibliography
fa A factor, given in table IV-a, that relates the maximum
process standard deviation to the difference between U
A bibliography of documents used in the devel
t of this
and L.
In ternational Standard is given in annex D.
k The acceptability constant when using the l ’s” method,
“a” method or “R” method.
K The acceptability constant when both ,u and 0 are 4 Acceptable Quality Level (AQL)
known.
4.1 Definition
L Lower specification limit. (As a
suffix to a variable,
denotes its value at L.)
When a continuous series of lots is considered, a quality level
U Upper specification limit. (As a suffix to a variable,
which for the purposes of sampling inspection is the limit of a
denotes its value at U.)
satisfactory process average percent nonconforming.
n Sample size (number of units in a sample).
4.2 Use
N Lot size (number of units in a lot).
P, The probability of acceptance.
The AQL, together with the sample size code letter, is used to
index the sampling plans in this International Standard.
Q The quality statistic.
Q, Lower quality statistic.
4.3 Specifying AQLs
QU Upper quality statistic.
The AQL to be used will be designated in the product specifica-
Standard deviation of a sample (estimate of the standard
s
tion contract or by the responsible authority. Where both upper
deviation of the process).
and lower specifications limits are given, separate AQLs may be
given to the individual limits, which are then known as
“separate double specification limits ”. Alternatively, an overall
AQL may be given which applies to the combined percent non-
conforming at both the upper and lower limits; this is then
known as a “combined double specification limit ”.
(See also annex A.)
4.4 Preferred AQLs
X Measured value of a characteristic in the sample.
Eleven AQLs given in this International Standard, ranging in
value from 0 ’10 % to 10 % nonconforming, are described as
X Mean value of x for the sample of n items.
preferred AQLs. If, for any product or service, an AQL is
designated other than a preferred AQL, then this International
Lower acceptance value.
Standard is not applicable. (See 12.2.)
F” Upper acceptance value.
Two further AQLs, 0,065 % and 15 %, are given to complete
the range of plans needed for the operation of the switching
Mean of the process.
lu
rules. (See clauses 19 and 21 .I Plans and curves designated by
an AQL of 0,065 % or 15 % will only be used when the AQL
g Standard deviation of the process. (~3, the square of the
under normal inspection is 0 ’10 % or 10 % respectively and the
standard deviation, is known as the variance.)
switching rules are invoked.
C “The sum of” (for example, C x = the sum of the x
values).
4.5 Caution
n
From the above definition of the AQL, it follows that desired
xi The sum of all the x values when i takes integral
protection can only be obtained when a continuous series of
z
values from 1 to n.
lots is provided for inspection,
i= 1
Is0 3951 : 1989 (E)
to the method chosen ( “s ”, “a” or, contingently, “R ”).
4.6 Limitation
Separate tables are given for normal, tightened and reduced
inspection.
The designation of an AQL shall not imply that the supplier has
the right to supply knowingly any nonconforming unit of
d) The switching rules are essentially the same.
product.
e) The classification of nonconformities by degree of
seriousness into class A, class B, etc. remains unchanged.
5 Switching rules for normal, tightened and
6.2 Differences
reduced inspection
a) Determination of acceptability. The acceptability of
an attributes sampling plan, taken from IS0 2859, is deter-
5.1 In order to discourage the producer from operating at a
process average percent nonconforming that exceeds the AQL, mined by the number of nonconforming units found in the
sample; the acceptability criterion in inspection by variables
this International Standard prescribes a switch to tightened in-
spection when inspection results indicate that the process is based on estimates of the location and variability of the
distributed measurements of the lot, in relation to the
average does exceed the AQL, and discontinuation of sampling
inspection altogether when tightened inspection does not in specification limits, that is in terms of the mean and stan-
dard deviation. In this International Standard two methods
time stimulate the producer to improve his production process.
are considered: the “s” method for use when the process
standard deviation 0 is unknown and the “a” method for
5.2 Hence, tightened inspection and the discontinuation rule
use when 0 is considered to be known. A third method,
are integral, and therefore obligatory, procedures of this Inter-
called the “R” method, is given in annex C. In the case of a
national Standard if the protection implied by the AQL is to be
single specification limit or of two separate limits, the ac-
maintained.
ceptability may be calculated from a formula (see 14.2 and
15.21, but is more easily established by a graphical method
5.3 This International Standard also provides the possibility of
(see 14.3). In the case of a combined double limit, this Inter-
switching to reduced inspection when inspection results indicate
national Standard provides for a graphical method (see 14.6
that the process average percent nonconforming is stable and
and 15.3).
reliable at a level below the AQL. This practice is, however, op-
b) Normality. In IS0 2859 there is no requirement
tional (at the discretion of the responsible authority).
relating to the distribution of the characteristics, but in this
International Standard it is necessary to the efficient opera-
5.4 When there is sufficient evidence from the control charts
tion of a plan that the measurements should be distributed
(see 18.1) that the variability is in statistical control, considera-
according to a normal distribution or a close approximation
tion should be given to switching to the “a” method. If this
to the normal distribution.
appears advantageous, the consistent value of s shall be taken
c) Operating characteristic curves (OC curves). While
as 0.
an individual variables plan may be devised the OC curve of
which corresponds closely to that of a given attributes plan,
5.5 When it has been necessary to discontinue sampling in-
it would not be possible to make all the OC curves in this In-
spection, inspection may not be resumed until action is taken by
ternational Standard identical with the corresponding OC
the producer to improve the quality of the submitted product.
curves in IS0 2859 (which are indexed with the same code
letter and AQL), without the sample size increasing with the
AQL for a given sample size code letter. For the l ’s” method
5.6 Details of the operation of the switching rules are given in
the sample size has been kept fixed for a given lot size
clauses 19 and 20.
across the full range of AQLs; for the “a” method this con-
straint has been removed in order to match the l ’s” and “o”
method OC curves as closely as possible, both at the AQL
6 Relation to IS0 2859
and at the limiting quality.
d) Probability of acceptance at the AQL. The prob-
6.1 Similarities
ability that a lot, whose quality is precisely at the AQL, will
be accepted increases with the sample size and follows a
a) This International Standard is a complement to
similar, but not identical, scale to that used in IS0 2859.
IS0 2859; the two documents share a common philosophy
The variables sample sizes corre-
and, as far as possible, their procedures and vocabulary are e) Sample sizes.
the same. sponding to given code letters are usually smaller than the
attributes sample sizes for the same letters.
b) Both use the AQL to index the sampling plans and the
sampling plans are
preferred values used in this document are identical with f) Double sampling plans. No double
those given in IS0 2859 for the same range of values (i.e. given in this International Standard.
from Of1 % to 10 %).
g) Average Outgoing Quality Limit (AOQL). Under
c) In the two International Standards, lot size and inspec- destructive or expensive testing, where 100 % inspection
tion level (inspection level II being preferred in default of and rectification of non-accepted lots is not feasible, the
other instructions) determine a code letter. Then general AOQL concept cannot be applied. As variables plans will
tables give the sample size to be taken and the acceptability generally be used under these circumstances, no AOQL
criterion in terms of the code letter and the AQL according tables have been included in this International Standard.
IS0 3951 : 1989 E)
7 Non-continuous production and operating 7.2 Operating characteristic curves
characteristic curves
a) The degree of protection of the consumer provided by the
individual sampling plans of this International Standard can,
7.1 Non-continuous production
however, be judged from their operating characteristic (OC)
curves as given in charts V-B to V-P and tables V-B-l to V-P-l,
a) The sampling schemes contained in this International
and these should be consulted when choosing a sampling plan.
Standard were not designed to be applied under cir-
b) These curves are for normal inspection using the ‘3”
cumstances different from those specified in 1.2, for
method with a single specification limit, but they provide a
example to an isolated lot or limited number of lots, where
good approximation to the case of a combined double
tightened inspection and the discontinuation rules cannot
specification limit.
be applied.
c) These curves also provide a good approximation to
b) Under such conditions, the concept of an AQL
the OC curves for the “a” method (and the “R” method)
becomes irrelevant, as the consumer ’s concern narrows to
sampling plans indexed by the same code letters and AQL,
the quality of the limited number of submitted lots, and he is
unless the sample size is small.
no longer involved with exerting controls on the quality of
the production process. The AQL will still indicate a submit- d) Separate OC curves are not given for tightened or re-
ted quality which has a high probability of acceptance, and duced inspection, but the curves may be found from among
can therefore still be used as an index to a sampling plan. those given for normal inspection (see tables VI-A
and VI-B).
(See clause 12.)
ISO 3951 :1989(E)
Section two: Choice of sampling plan
NOTES
8 Planning
1 Tests for departure from normality are dealt with in section two of
The choice of the most suitable variables plan, if one exists, re-
IS0 2854, which provides examples of graphical methods which can
be used to verify that the distribution of the data is sufficiently normal
quires experience, judgement and some knowledge of both
to justify the use of sampling by variables.
statistics and the product to be inspected. This section of this
International Standard is intended to suggest to those
2 IS0 5479 gives a more detailed study on the subject of normality
responsible for specifying sampling plans the considerations
tests.
that should be borne in mind when deciding whether a
variables plan would be suitable and the choices to be made
when selecting an appropriate standard plan.
10 Choice of method
If it is desired to apply inspection by variables, the next ques-
tion is which method should be used, the “s” method or the
9 Choice between variables and attributes
“a” method (or the “R” method).
The first question to consider is whether it is desirable to
The “a” method is the most economical in sample size, but
inspect by variables rather than by attributes. The following
before this method may be employed, the value of 0 has to be
points should be taken into account :
established.
a) In terms of economics, it is necessary to compare the
In terms of sample size, the “s” method has a slight advantage
total cost of the relatively simple inspection of a larger
over the “R” method, but the calculation of s involves more
number of items by an attributes scheme with the generally
computation; the extent and difficulty of this is more apparent
more elaborate procedure required by a variables scheme,
than real, especially if an electronic calculator is available.
which is usually more expensive in time and money per
Methods of calculating s are given in annex A.
item.
The “R” method (given in annex C) is simple to calculate, but
b) In terms of the knowledge gained, the advantage lies
with inspection by variables as the more precise information requires a somewhat larger sample size for the same AQL. Also
it has the undesirable property that for samples of size 10 or
obtained indicates how good the product is; earlier warning
will be given if the quality is slipping. more the acceptability of a lot can depend upon how the
sample is divided into sub-groups.
c) An attributes scheme can be more readily understood
and accepted; for example, it may at first be difficult to ac-
Initially, it will be necessary to begin with the “s” (or the
cept that, when inspecting by variables, a lot can be re-
“R” method), but if the quality is satisfactory, the standard
jected on measurements taken of a sample that does not
switching rules will permit the responsible authority to com-
contain any nonconforming items. (See the example
mence reduced inspection and use a smaller sample size.
in 14.6.)
The question then is, if the variability is under control and lots
d)
A comparison of the size of the samples required for the
continue to be accepted, will it be economical to change to the
same AQL from standard plans for inspection by attributes
“a” method ?
(i.e. from IS0 2859) and the standard plans in this Inter-
national Standard is given in table I-B. It will be seen that
The size of the sample will generally be smaller and the accept-
the smallest samples are required by the “a” method, used
ability criteria become simpler. (See clauses 15.2 and 15.3.) On the
when the standard deviation of the process is known.
other hand, it will still be necessary to calculate s for record pur-
poses and to keep the control charts up to date. (See clause 18.1
e) Inspection by variables is appropriate particularly in
conjunction with the use of control charts for variables.
f) Variables sampling has a substantial advantage when
11 Choice of inspection level and AQL
the inspection process is expensive, for example in the case
of destructive testing.
In standard sampling plans, the inspection level in conjunction
with the size of the lots and the AQL determines the size of the
g) A variables scheme becomes less suitable as the
sample to be taken, and governs the severity of the inspection.
number of measurements to be taken on one item in-
The appropriate OC curve given in one of the tables V-B to V-P
creases, as each characteristic has to be considered
shows the extent of the risk that is involved in such a plan.
separately. It may be advantageous to apply “attributes” to
the majority of the characteristics and “variables” to one or
The choice of the inspection level and AQL is governed by a
two of the more important requirements, for example proof
number of factors, but is mainly a balance between the total
load tests, safety and reliability requirements.
cost of inspection and the consequences of nonconforming
h) The use of this International Standard is only applicable items passing into service.
when there is reason to believe that the distribution of
measurements is normal. In case of any doubt, the respon-
The normal practice is to use inspection level II, unless special
sible authority should be consulted. circumstances indicate that another level is more appropriate.
IS0 3951 : 1989 (E)
12 Choice of sampling plan this International Standard may be selected by using
diagram A. The intersection of a vertical line through the
acceptable value for the limiting quality and a horizontal
12.1 Standard plans
line through the desired quality with a 95 % probability of
acceptance (approximately equal to AQL) will lie on,or under, a
The standard procedure can be used only when the production
sloping line indexed with the sample size code letter of a stan-
of lots is continuous.
dard plan which meets the specified requirements. This should
be verified by inspecting the OC curve given in table V for this
The standard procedure, with its semi-automatic steps from lot
code letter and AQL.
size to sample size, using inspection level II and beginning with
the “s” method, has been found in practice to produce
If the lines intersect at a point above the line marked P (see
workable sampling plans; but it assumes that the order of
diagram A), this implies that, for example, a sample of over 200
priority is first the AQL, second the sample size and last, the
would be necessary for the s method and the specification can-
limiting quality.
not be met by the plans in this International Standard.
The acceptability of this system is due to the fact that the con-
sumer is protected by the switching rules (see clause 191,
12.2 Special plans
which quickly increase the severity of inspection and finally ter-
minate it, if the quality of the process is worse than the AQL.
If standard plans are not acceptable, it will be necessary to
devise a special plan. The choice is then to decide which com-
NOTE - It should also be remembered that the limiting quality is the
bination of AQL, limiting quality, and sample size is most
quality which if offered for inspection would have a 10 % probability of
suitable, remembering that these are not independent, for,
acceptance. The actual risk taken by the consumer therefore also
depends on the probability of goods of this low quality being offered
when any two have been chosen, the third follows.
for inspection.
NOTE - This choice is not completely unfettered; the fact that the size
However, if, in certain circumstances, the limiting quality has a
of the sample is necessarily a whole number imposes some constraints.
higher priority than the sample size (for example, when only a
If a special scheme is necessary it should be devised only with the
limited number of lots are being produced), a suitable plan in assistance of a statistician experienced in quality control.
IS0 3951 :1989(E)
Section three: Operation of a variables sampling plan
Thus, if only the upper specification limit U is given, the lot is
13 Preliminary operations
Before starting inspection by variables, check
acceptable if 0” > k
a) that the distribution can be considered to be normal and
not acceptable if Q” < k
that production is considered to be continuous;
b) whether the “s” (or “R ”) method is to be used initially
Or, if only the lower specification limit L is given, the lot is
or whether the standard deviation is stable and known, in
which case the “a” method should be used;
acceptable if QL > k
c) that the inspection level to be used has been
not acceptable if QL < k
designated. If none has been given, inspection level II shall
be used;
When both U and L are given (k values are different if the AQLs
d) that the AQL has been designated and that it is one of
are different for the upper limit and the lower limit), the lot is
the preferred AQLs for use with this International Standard.
If it is not, the tables are not applicable;
acceptable if both 0, > kL and Q” > kU
e) if a double specification limit has to be met, whether the
not acceptable if either QL < kL or QU < k”
limits are separate or combined and, if the limits are
separate, whether AQLs are determined for each limit.
Example
The maximum temperature of operation for a certain device is
14 Standard procedure for “sl’ method
specified as 60 OC. Production is inspected in lots of 100 items.
Inspection level II, normal inspection with AQL = 2 ’5 % is to
14.1 Obtaining a plan
be used. From table I-A, the sample size code letter is F; from
table II-A it is seen that a sample size of 10 is required and that
The procedure for obtaining a plan is as follows :
the acceptability constant k is 1 ’41. Suppose the measure-
a) With the inspection level given (normally this will be II) ments are as follows: 53 OC; 57 “C; 49 OC; 58 “C; 59 “C; 54 OC;
and with the lot size, obtain the sample size code letter 58 OC; 56 OC; 55 OC; 50 OC. Compliance with the acceptability
using table I-A.
criterion is to be determined.
b) With this code letter and the AQL, enter table II-A and
obtain the sample size n and acceptability constant k.
Information needed Values obtained
c) Taking a random sample of this size, measure the
Sample size : n 10
characteristic x in each item and then calculate X, the
sample mean, and s, the estimated standard deviation (see
Sample mean X: C x/n
w9
annex A). lf X is outside the specification limit, the lot can
be judged unacceptable without calculating s. It may,
Standard deviation s : J C (xi - FP/(n - 1) 3,414
however, be necessary to calculate s for record purposes.
(See A.l.2, annex A.) ’
14.2 Acceptability criteria for single or separate
Specification limit (upper) : U 60
double specification limits
QU = (U - X)/s 1,494
If single or separate specification limits are given, calculate the
quality statistic
Acceptability constant : k (see table II-A) 1 ’41
u-x
QU= -
Acceptability criterion : compare 0, with k 1,494 > 1 ’41
S
and/or The lot meets the acceptability criterion, and is therefore accep-
table.
X-L
QL = -
S
14.3 Graphical method for a single specification
as appropriate,
limit
then compare the quality statistic (Q, and/or Q,) with the
When a graphical criterion is desired, draw the line
acceptability constant k obtained from table II-A for normal
inspection. If the appropriate quality statistic is greater than or
j+
U - k s (for an upper limit) or
equal to the acceptability constant, the lot is acceptable; if less,
it is not acceptable. X = L + k s (for a lower limit),
IS03951 :I989 (I3
as appropriate, on graph paper with Xas the vertical axis and s stants are k” = I,57 and kL = 254 respectively. Suppose the
as the horizontal axis. When the inspection concerns an upper sample delay times are as follows:
specification limit the accept zone is the zone below the line.
6,95 6,65
When a lower specification limit is considered, the accept zone 6 ’04
6 ’68 663
is the zone above the line. Using the values of s and X 6,34 6,15
6 ’40 6 ’4 604
7,15 6,70 6,51
calculated from the measurements obtained from a sample (see 644 659
7,17
annex A for the calculation of s), plot the point (s, XI on the 6,35 6,25 6,96
W33
graph. If this point lies in the accept zone, the lot is acceptable; 6,80 6,15 6,25 6,57
if outside, it is not acceptable. 6,52 6,59 6,86 6,57 6,91
6,29 6,67 6,67
663 6,70
Example
Compliance with the acceptability criteria is to be determined.
Using the data given in the example in 14.2, mark the point
u= 60 on the X (vertical) axis and draw a line through this
Information needed Value obtained
point with a slope - k [as k = 1,41, this means the line passes
through points (s = 1, X = 58,591, (s = 2, X = 57,18), (s = 3,
Sample size: n 35
X=
55,771, etc.]. Select a suitable point and draw a straight
line through it and (s = 0, X = 60), i.e. U. The accept zone is
Sample mean X: C x/n 6,55 s
then the area under this line. The calculated values of s and X
are 3,414 and 54,9. Plotting the point (s, X), it will be seen from
Sample standard deviation s :
figure 1 that it lies just inside the accept zone; the lot is accep-
- XP/(n - I)
-\/c (Xi 0,31 s
table.
(See A.l.2, annex A.)
The graph can be prepared before beginning the inspection of a
series of lots. Then, for each lot plot the point (s, X) and decide
Upper specification limit: U 9,0 s
whether or not the lot is acceptable.
QU = (U - X)/s
7 ’90
z U=60
t k =I,41
Acceptability constant: ku (see table II-A) I,57
Lower specification limit: L 4,0 s
-
QL = (x - L)ls 8,23
zone
56 -
Acceptability constant: kL (see table II-A)
2 ’54
Acceptability criterion: is QU > kU 7,90> I,57
and QL > kL?
and 8,23 > 254
The lot meets the acceptability criteria, and is acceptable.
t-lgure 1 - Example of the use of an acceptance chart
for a single speci f ication limit : 3” method
14.5 Graphical method for separate double
specification limits
14.4 Numerical method for single or separate
double specification limi
ts
When a graphical criterion is desi red for separate double
specification limits, draw the lines
Example
ji-=
U - k ”s (for the upper limit) and
A certain pyrotechnic delay has a specified minimum delay time
of 4,0 s and a maximum time of 9,0 s. Production is inspected
X = L + kL s (for the lower limit)
in lots of 1 000 items and inspection level II, normal inspection,
is to be used with an AQL of 0,l % applied to the lower limit on graph paper with Xas the vertical axis and s as the horizontal
and an AQL of 2,5 % applied to the upper limit. From table I-A axis. Using the values of s and x calculated from the
it is seen that the sample size code letter is J; from table I-B it is measurements obtained from a sample, plot the point (s, X) on
seen that the sample size is 35 for the “s” method and from the graph. If this point lies in the accept zone, the lot is accep-
table II-A it is found that the upper and lower acceptability con- table; if outside, it is not acceptable.
IS0 3951 :1989 (E)
and tightened inspection be copied (or traced) onto graph
Example
paper. (A sheet of the special graph paper needed for this is
given in annex D.) The scales should be adjusted so that s and
Using the data given in the example in 14.4 mark the point
X can be plotted directly (e.g. the upper limit is given
u= 9,O on the x (vertical) axis and draw a line through this
instead of I,0 and the lower limit instead of 0 on the STscale).
point with a slope - kU [as kU = 1 ’57, this means that the line
will pass through the point (s = 1, X = 7/W]. Also mark the
Then plot on the chart the values of s and xfound from the
point L = 4,O on the Taxis and draw a line through this point
sample and, if the point lies inside the accept zone, the lot is ac-
with slope + kL [as kL = 254, this means that the line will pass
ceptable; if outside, it is not acceptable.
through the point (s = 1, X = 6,54)]. The accept zone is then
the area bounded by these lines and the Kaxis. The calculated
-
NOTE - For sample size code letters B and C (i.e. sample sizes 3 and
values of s and x are 0 ’31 and 6 ’55. Plotting the point
4), the accept zone is bounded by four straight lines: the Taxis, the line
-
b = 0 ’31, X = 6 ’55) as shown in figure 2 it will be seen that it
x= u - k s, a line parallel to the x axis through the MSSD (see
lies well within the accept zone and that the lot is acceptable.
table IV-s) and the line x = L + ks. The value of k is obtained from
the table II-A, II-B or II-C.
Example
The minimum temperature of operation for a certain device is
specified as 60 ’0 OC and the maximum temperature as 70 ’0 OC.
Production is in inspection lots of 96 items. Inspection level II,
normal inspection, with AQL = I,5 %, is to be used. From
table I-A, the sample size code letter is F; from table I-B it is
seen that a sample size of 10 is required and from table IV-s
that the value of fs for the MSSD is 0,276. Suppose the
measurements obtained are as follows : 63 ’5 OC; 62 ’0 OC;
65 ’2 OC; 61 ’7 OC; 69 ’0 “C; 67 ’1 “C; 60 ’0 OC; 66 ’4 OC; 62 ’8 OC;
68 ’0 OC. Compliance with the acceptability criterion is to be
determined.
6-
Information needed Value obtained
Sample size
...
SLOVENSKI STANDARD
01-september-1996
Sampling procedures and charts for inspection by variables for percent
nonconforming
Sampling procedures and charts for inspection by variables for percent nonconforming
Règles et tables d'échantillonnage pour les contrôles par mesures des pourcentages de
non conformes
Ta slovenski standard je istoveten z: ISO 3951:1989
ICS:
03.120.30 8SRUDEDVWDWLVWLþQLKPHWRG Application of statistical
methods
2003-01.Slovenski inštitut za standardizacijo. Razmnoževanje celote ali delov tega standarda ni dovoljeno.
INTERNATIONAL Is0
STANDARD 3951
Second edition
1989-09-15
Sampling- procedures and charts for inspection
by variables for percent nonconforming
R&g/es et tables d ’khan tiilonnage pour /es contr&es par mesures des pourcentages
de non con formes
Reference number
IS0 3951 : 1989 (El
IS0 3951 : 1989 (El
Foreword
IS0 (the International Organization for Standardization) is a worldwide federation of
national standards bodies (IS0 member bodies). The work of preparing International
Standards is normally carried out through IS0 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, govern-
mental and non-governmental, in liaison with ISO, also take part in the work. IS0
collaborates closely with the International Electrotechnical Commission (IEC) on all
matters of electrotechnical standardization.
Draft International Standards adopted by the technical committees are circulated to
the member bodies for approval before their acceptance as International Standards by
the IS0 Council. They are approved in accordance with IS0 procedures requiring at
least 75 % approval by the member bodies voting.
International Standard IS0 3951 was prepared by Technical Committee ISO/TC 69,
Applications of stat/s tical methods.
This second edition cancels and replaces the first edition (IS0 3951 : 1981), of which it
constitutes a technical revision.
The principal changes to the first edition are as follows:
a) a distinction has been drawn between maximum process standard deviation
(MPSD) under the “a” method and maximum sample standard deviation (MSSD)
under the l ’s” method;
b) all the “a” method acceptance curves have been truncated at the appropriate
maximum process standard deviations;
c) the terminology has been aligned on that of IS0 2859 and IS0 3534.
Users should note that all International Standards undergo revision from time to time
and that any reference made herein to any other International Standard implies its
latest edition, unless otherwise stated.
0 IS0 1989
All rights reserved. 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 the publisher.
International Organization for Standardization
Case postale 56 l CH-1211 Geneve 20 l Switzerland
Printed in Switzerland
ii
IS0 3951 : 1989 (E)
Contents
Page
Section one : General
.................................................
1 Scope and field of application
........................................
........................................................
2 References
3 Definitions and symbols.
............................................
4 Acceptable Quality Level (AQL) . ; .
5 Switching rules for normal, tightened and reduced inspection. .
6 Relation to IS0 2959.
...............................................
7 Non-continuous production and operating characteristic curves. .
Section two : Choice of sampling plan
.................................
8 Planning .
9 Choice between variables and attributes
...............................
10 Choice of method
..................................................
II Choice of inspection level and AQL
...................................
12 Choice of sampling plan
.............................................
Section three : Operation of a variables sampling plan
.................. 8
13 Preliminary operations 8
..............................................
14 Standard procedure for “s” method.
.................................. 8
15 Standard procedure for “0” method
.................................. 12
Procedure during continuing inspection
16 . 15
17 Normality and outliers
.............................................. 15
Records .
18 15
19 Operation of switching rules
......................................... 16
Discontination and resumption of inspection
20 . 16
21 Acceptance curves for tightened and reduced inspection.
................ 16
Switching to “a” method
22 . 16
. . .
III
IS0 3951 : 1989 (E)
Page
Section four : Tables and diagrams . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
Tables
I-A Sample size code letters and inspection levels . . . . . . . . . . . . . . 18
I-B Sample size code letters and sample sizes for normal
inspection. 19
Single sampling plans for normal inspection
II-A
(master table) : l ’s” method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
II-B Single sampling plans for tightened inspection
(master table) : “s” method . . . . . . . . i . . . . . . . . . 21
Single sampling plans for reduced inspection
II-C
(master table) : l ’s” method. . . . . . . . . . . . . . . . . . 22
Ill-A Single sampling plans for normal inspection
(master table) : “a” method . . . . . . . . . . . . . . . . . 23
Single sampling plans for tightened inspection
III-B
(master table) : “a” method . . . . . . . . . . . . . . . . . 24
III-C Single sampling plans for reduced inspection
(master table) : “a” method . . . . . . . . . . . . . . . . . . . 25
Values of “fS,, for maximum sample standard
IV-S
deviation (MSSD) : “s ”method . . . . . . . . . . . . . . 26
IV-0 Values of “f0,’ for maximum process standard
deviation (MPSD) : “a” method . . . . . . . . . . . . . . . . . . . . . . . . 26
V (V-B to V-P) Tables and operating characteristic curves for single
sampling plans (sample size code letters B to P) 27-40
Correspondence between tig htened and normal
VI-A
inspection plans. 41
VI-B Correspondence between reduced and normal
inspection plans.
Diagrams
Sample size code letters of standard sampling plans for specified
A
qualities at 95 % and 10 % probabilities of acceptance . . . . . . .
s-D to s-P Acceptance curves for combined double specification
limits : ‘Is” method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43-54
Acceptance curves for combined double specification
a-c to a-P
limits : “a” method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55-67
Annexes
.....................................
A Procedures for obtaining s and 0 68
..................................................
B Statistical theory. 71
C Sampling plans for “R” method . 82
......................................
D Graph paper for the I ’s” method 105
Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107
IV
INTERNATIONAL STANDARD
IS03951 : 1989 (E)
Sampling procedures and charts for inspection by
variables for percent nonconforming
Section one : General
normal distribution or a close approximation to the normal
1 Scope and field of application
distribution;
d) where a contract or standard defines an upper
1.1 Scope
specification limit U, a lower specification limit L, or
both; a product is qualified as nonconforming when its
1 .I .I This International Standard establishes sampling plans
measured quality characteristic x satisfies one of the fol-
and procedures for inspection by variables. It is complementary
lowing inequalities :
to IS0 2859. When specified by the responsible authority, both
x>u . . . (1)
this International Standard and IS0 2859 may be referenced in
a product or process specification contract, inspection instruc-
x
tions’ or other documents and the provisions set forth therein
eitherx > Uorx < L
. . . (3)
shall govern. The “responsible authority” shall be designated in
one of the above documents.
Inequalities (1) and (2) are called cases with a single specifica-
tion limit, and (3) a case with double specification limits. In
this last situation a further distinction is made between separate
1.1.2 The object of the methods laid down in this Inter-
or combined double limits according to whether the AQL is
national Standard is to ensure that lots of an acceptable quality
applied to each limit separately or to both limits combined (see
have a high probability of acceptance and that the probability
clause 4).
of not accepting inferior lots is as high as possible.
1.1.3 In common with IS0 2859, the percentage of noncon-
2 References
forming products in the lots is used to define the quality of
these lots and of the production process in question. I SO 2854, Statistical interpretation of data - Techniques of
estimation and tests relating to means and variances.
1.2 Field of application IS0 2859, Sampling procedures and tables for inspection by
attributes.
This International Standard is primarily designed for use under
IS0 3534, Statistics - Vocabulary and symbols.
the following conditions :
I SO 5479, Normality tests. 1)
a) where the inspection procedure is to be applied to a
IS0 5725, Precision of test methods - Determination of
continuous series of lots of discrete products all supplied
repeatability and reproducibility for a standard test method by
by one producer using one production process. If there are
inter-laboratory tests.
different producers this International Standard shall be ap-
plied to each one separately;
b) where only a single quality characteristic x of these
3 Definitions and symbols
products is taken into consideration, which must be
measurable on a continuous scale. If several such
3.1 Definitions
characteristics are of importance, this International Stan-
dard shall be applied to each separately;
For the purposes of this International Standard, the definitions
c) where production is stable (under statistical control) given in IS0 3534 and IS0 2859, together with the following,
and the quality characteristic x is distributed according to a
apply.
1) At present at the stage of draft.
IS0 3951 :I989 (E)
3.1 .I0 specification limit (limite de specification) : The
3.1.1 inspection by variables (controle par mesures) : A
limiting value (lower or upper) specified for a quantitative
method which consists in measuring a quantitative character-
characteristic.
istic for each item of a population or a sample taken from
this population.
3.1 .I1 lower specification limit (L) [limite inferieure de spe-
cification (Li)]: A specification limit that defines the lower con-
3.1.2 acceptance sampling by variables (echantillonnage
formance boundary for an individual unit of a manufacturing or
par mesures en vue d ’acceptation) : An acceptance procedure
service operation.
wherein a specified characteristic is measured to establish
statistically the acceptability of a lot from the result obtained
3.1.12 upper specification limit (U) [limite superieure de
from the items in a sample.
specification (L,)]: A specification limit that defines the upper
conformance boundary for an individual unit of a manufactur-
3.1.3 acceptable quality level (AQL) [niveau de qualite ac-
ing or service operation.
ceptable (NQA)] : When a continuous series of lots is con-
sidered, a quality level which for the purposes of sampling in-
n limit (limite unique de specifica-
3.1.13 single specificatio
spection is the limit of a satisfactory process average percent
tion) : The term used when one limit only is specified.
nonconforming. (See clause 4.)
3.1.14 separate double specification limits (limites de
3.1.4 limiting quality (qualite limite) : When a lot is con-
specifications doubles separees) : The term used when both
sidered in isolation, a quality level which for the purposes of
upper and lower limits are specified and separate AQLs are
sampling inspection is limited to a low probability of accep-
applied to each limit individually. (See 4.3.)
tance (in this International Standard : 10 %). (See 12.1.)
3.1.15 combined double specification limit (limite de
specification double combinee) : The term used when both
3.1.5 nonconformity (non-conformite) : Failure to fulfil a
specified requirement by a quality characteristic of an item or upper and lower limits are specified and an AQL is given which
service, the assessment of which does not depend essentially applies to the combined percent nonconforming at the two
on the passage of time. limits. (See 4.3.)
Nonconformities will generally be classif ied by their degree of
3.1 .I6 acceptability constant (k) [constante d ’acceptabilite
seriousness, such as
(HI : A constant dependent on the specified value of the ac-
ceptable quality level and the sample size. (See 14.2 and 15.2,
or clause C.5 in annex C.)
Class A. Those nonconformities of a type considered to be of
the highest concern for the product or service. In acceptance
sampling, such types of nonconformity will be assigned very 3.1.17 quality statistic (Q) [statistique de qualite (Q)] : A
small AQL values. function of the specification limit, the sample mean, and the
standard deviation. The lot is sentenced on the result of com-
paring Q with the acceptability constant k. (See 14.2 and 15.2,
Class B. Those nonconformities of a type considered to have
or clause C.5 in annex C.)
the next lower degree of concern; therefore these can be
assigned a larger AQL value than those in class A and smaller
3.1.18 lower quality statistic (0,) [statistique de qualite
than in class C, if a third class exists, and so on.
correspondant a la limite inferieure (Qi)] : A function of the
lower specification limit, the sample mean, and the standard
The number of classes and the assignment into a class should
deviation. The lot is sentenced on the result of comparing Q,
be appropriate to the quality requirements of the specific situa-
with the acceptability constant k. (See 14.2 and 15.2, or clause
tion.
C.5 in annex C.)
3.1.6 nonconforming u nit (unite non conforme): A unit of
3.1.19 upper quality statistic (QJ [statistique de qualite
product or service contain ing at least one nonconformity.
correspondant a la limite superieure (QJ] : A function of the
upper specification limit, the sample mean, and the standard
3.1.7 “s” method (methode ((SD) : A method of assessing the
deviation. The lot is sentenced on the result of comparing QU
acceptability of a lot by using the sample standard deviation.
with the acceptability constant k. (See 14.2 and 15.2, or clause
(See clause 14.1
C.5 in annex C.)
3.1.8 “o“ method (methode ((0))) : A method of assessing
3.1.20 maximum sample standard deviation (MSSD)
the acceptability of a lot by using knowledge of the process
[&art-type maximal de I ’echantillon (ETME)]: Under given con-
standard deviation, (See clause 15.)
ditions, the largest acceptable sample standard deviation. (See
14.6 and B.8.3 in annex B.)
3.1.9 “R” method (methode
the acceptability of a lot by indirectly using an estimate of the
3.1.21 maximum process standard deviation (MPSD)
standard deviation of the process based on the average range [&art-type maximal du procede (ETMP)]: Under given con-
of the measurements of the items in sub-groups of the sample.
ditions, the largest acceptable process standard deviation. (See
(See annex C.) 15.3 and B.5.2 in annex B).
IS0 3951 : 1989 (E)
> “Greater than” (for example, a > b means a is greater
3.122 switching rules (regles de modification du controle) :
than 6).
Instructions within a sampling scheme for shifting from one
sampling plan to another based on demonstrated quality
“Greater than or equal to” (for example, a > b means a
history. (See clause 19.) >
is greater than or equal to b).
3.2 Symbols
< “Less than” (for example, a < b means a is less than
b).
The symbols used are as follows:
“Less than or equal to” (for example, a < b
G a is
, ‘s A factor, given in table IV-s, that relates the maximum
less than or equal to 6).
sample standard deviation to the difference between U
and L.
3.3 Bibliography
fa A factor, given in table IV-a, that relates the maximum
process standard deviation to the difference between U
A bibliography of documents used in the devel
t of this
and L.
In ternational Standard is given in annex D.
k The acceptability constant when using the l ’s” method,
“a” method or “R” method.
K The acceptability constant when both ,u and 0 are 4 Acceptable Quality Level (AQL)
known.
4.1 Definition
L Lower specification limit. (As a
suffix to a variable,
denotes its value at L.)
When a continuous series of lots is considered, a quality level
U Upper specification limit. (As a suffix to a variable,
which for the purposes of sampling inspection is the limit of a
denotes its value at U.)
satisfactory process average percent nonconforming.
n Sample size (number of units in a sample).
4.2 Use
N Lot size (number of units in a lot).
P, The probability of acceptance.
The AQL, together with the sample size code letter, is used to
index the sampling plans in this International Standard.
Q The quality statistic.
Q, Lower quality statistic.
4.3 Specifying AQLs
QU Upper quality statistic.
The AQL to be used will be designated in the product specifica-
Standard deviation of a sample (estimate of the standard
s
tion contract or by the responsible authority. Where both upper
deviation of the process).
and lower specifications limits are given, separate AQLs may be
given to the individual limits, which are then known as
“separate double specification limits ”. Alternatively, an overall
AQL may be given which applies to the combined percent non-
conforming at both the upper and lower limits; this is then
known as a “combined double specification limit ”.
(See also annex A.)
4.4 Preferred AQLs
X Measured value of a characteristic in the sample.
Eleven AQLs given in this International Standard, ranging in
value from 0 ’10 % to 10 % nonconforming, are described as
X Mean value of x for the sample of n items.
preferred AQLs. If, for any product or service, an AQL is
designated other than a preferred AQL, then this International
Lower acceptance value.
Standard is not applicable. (See 12.2.)
F” Upper acceptance value.
Two further AQLs, 0,065 % and 15 %, are given to complete
the range of plans needed for the operation of the switching
Mean of the process.
lu
rules. (See clauses 19 and 21 .I Plans and curves designated by
an AQL of 0,065 % or 15 % will only be used when the AQL
g Standard deviation of the process. (~3, the square of the
under normal inspection is 0 ’10 % or 10 % respectively and the
standard deviation, is known as the variance.)
switching rules are invoked.
C “The sum of” (for example, C x = the sum of the x
values).
4.5 Caution
n
From the above definition of the AQL, it follows that desired
xi The sum of all the x values when i takes integral
protection can only be obtained when a continuous series of
z
values from 1 to n.
lots is provided for inspection,
i= 1
Is0 3951 : 1989 (E)
to the method chosen ( “s ”, “a” or, contingently, “R ”).
4.6 Limitation
Separate tables are given for normal, tightened and reduced
inspection.
The designation of an AQL shall not imply that the supplier has
the right to supply knowingly any nonconforming unit of
d) The switching rules are essentially the same.
product.
e) The classification of nonconformities by degree of
seriousness into class A, class B, etc. remains unchanged.
5 Switching rules for normal, tightened and
6.2 Differences
reduced inspection
a) Determination of acceptability. The acceptability of
an attributes sampling plan, taken from IS0 2859, is deter-
5.1 In order to discourage the producer from operating at a
process average percent nonconforming that exceeds the AQL, mined by the number of nonconforming units found in the
sample; the acceptability criterion in inspection by variables
this International Standard prescribes a switch to tightened in-
spection when inspection results indicate that the process is based on estimates of the location and variability of the
distributed measurements of the lot, in relation to the
average does exceed the AQL, and discontinuation of sampling
inspection altogether when tightened inspection does not in specification limits, that is in terms of the mean and stan-
dard deviation. In this International Standard two methods
time stimulate the producer to improve his production process.
are considered: the “s” method for use when the process
standard deviation 0 is unknown and the “a” method for
5.2 Hence, tightened inspection and the discontinuation rule
use when 0 is considered to be known. A third method,
are integral, and therefore obligatory, procedures of this Inter-
called the “R” method, is given in annex C. In the case of a
national Standard if the protection implied by the AQL is to be
single specification limit or of two separate limits, the ac-
maintained.
ceptability may be calculated from a formula (see 14.2 and
15.21, but is more easily established by a graphical method
5.3 This International Standard also provides the possibility of
(see 14.3). In the case of a combined double limit, this Inter-
switching to reduced inspection when inspection results indicate
national Standard provides for a graphical method (see 14.6
that the process average percent nonconforming is stable and
and 15.3).
reliable at a level below the AQL. This practice is, however, op-
b) Normality. In IS0 2859 there is no requirement
tional (at the discretion of the responsible authority).
relating to the distribution of the characteristics, but in this
International Standard it is necessary to the efficient opera-
5.4 When there is sufficient evidence from the control charts
tion of a plan that the measurements should be distributed
(see 18.1) that the variability is in statistical control, considera-
according to a normal distribution or a close approximation
tion should be given to switching to the “a” method. If this
to the normal distribution.
appears advantageous, the consistent value of s shall be taken
c) Operating characteristic curves (OC curves). While
as 0.
an individual variables plan may be devised the OC curve of
which corresponds closely to that of a given attributes plan,
5.5 When it has been necessary to discontinue sampling in-
it would not be possible to make all the OC curves in this In-
spection, inspection may not be resumed until action is taken by
ternational Standard identical with the corresponding OC
the producer to improve the quality of the submitted product.
curves in IS0 2859 (which are indexed with the same code
letter and AQL), without the sample size increasing with the
AQL for a given sample size code letter. For the l ’s” method
5.6 Details of the operation of the switching rules are given in
the sample size has been kept fixed for a given lot size
clauses 19 and 20.
across the full range of AQLs; for the “a” method this con-
straint has been removed in order to match the l ’s” and “o”
method OC curves as closely as possible, both at the AQL
6 Relation to IS0 2859
and at the limiting quality.
d) Probability of acceptance at the AQL. The prob-
6.1 Similarities
ability that a lot, whose quality is precisely at the AQL, will
be accepted increases with the sample size and follows a
a) This International Standard is a complement to
similar, but not identical, scale to that used in IS0 2859.
IS0 2859; the two documents share a common philosophy
The variables sample sizes corre-
and, as far as possible, their procedures and vocabulary are e) Sample sizes.
the same. sponding to given code letters are usually smaller than the
attributes sample sizes for the same letters.
b) Both use the AQL to index the sampling plans and the
sampling plans are
preferred values used in this document are identical with f) Double sampling plans. No double
those given in IS0 2859 for the same range of values (i.e. given in this International Standard.
from Of1 % to 10 %).
g) Average Outgoing Quality Limit (AOQL). Under
c) In the two International Standards, lot size and inspec- destructive or expensive testing, where 100 % inspection
tion level (inspection level II being preferred in default of and rectification of non-accepted lots is not feasible, the
other instructions) determine a code letter. Then general AOQL concept cannot be applied. As variables plans will
tables give the sample size to be taken and the acceptability generally be used under these circumstances, no AOQL
criterion in terms of the code letter and the AQL according tables have been included in this International Standard.
IS0 3951 : 1989 E)
7 Non-continuous production and operating 7.2 Operating characteristic curves
characteristic curves
a) The degree of protection of the consumer provided by the
individual sampling plans of this International Standard can,
7.1 Non-continuous production
however, be judged from their operating characteristic (OC)
curves as given in charts V-B to V-P and tables V-B-l to V-P-l,
a) The sampling schemes contained in this International
and these should be consulted when choosing a sampling plan.
Standard were not designed to be applied under cir-
b) These curves are for normal inspection using the ‘3”
cumstances different from those specified in 1.2, for
method with a single specification limit, but they provide a
example to an isolated lot or limited number of lots, where
good approximation to the case of a combined double
tightened inspection and the discontinuation rules cannot
specification limit.
be applied.
c) These curves also provide a good approximation to
b) Under such conditions, the concept of an AQL
the OC curves for the “a” method (and the “R” method)
becomes irrelevant, as the consumer ’s concern narrows to
sampling plans indexed by the same code letters and AQL,
the quality of the limited number of submitted lots, and he is
unless the sample size is small.
no longer involved with exerting controls on the quality of
the production process. The AQL will still indicate a submit- d) Separate OC curves are not given for tightened or re-
ted quality which has a high probability of acceptance, and duced inspection, but the curves may be found from among
can therefore still be used as an index to a sampling plan. those given for normal inspection (see tables VI-A
and VI-B).
(See clause 12.)
ISO 3951 :1989(E)
Section two: Choice of sampling plan
NOTES
8 Planning
1 Tests for departure from normality are dealt with in section two of
The choice of the most suitable variables plan, if one exists, re-
IS0 2854, which provides examples of graphical methods which can
be used to verify that the distribution of the data is sufficiently normal
quires experience, judgement and some knowledge of both
to justify the use of sampling by variables.
statistics and the product to be inspected. This section of this
International Standard is intended to suggest to those
2 IS0 5479 gives a more detailed study on the subject of normality
responsible for specifying sampling plans the considerations
tests.
that should be borne in mind when deciding whether a
variables plan would be suitable and the choices to be made
when selecting an appropriate standard plan.
10 Choice of method
If it is desired to apply inspection by variables, the next ques-
tion is which method should be used, the “s” method or the
9 Choice between variables and attributes
“a” method (or the “R” method).
The first question to consider is whether it is desirable to
The “a” method is the most economical in sample size, but
inspect by variables rather than by attributes. The following
before this method may be employed, the value of 0 has to be
points should be taken into account :
established.
a) In terms of economics, it is necessary to compare the
In terms of sample size, the “s” method has a slight advantage
total cost of the relatively simple inspection of a larger
over the “R” method, but the calculation of s involves more
number of items by an attributes scheme with the generally
computation; the extent and difficulty of this is more apparent
more elaborate procedure required by a variables scheme,
than real, especially if an electronic calculator is available.
which is usually more expensive in time and money per
Methods of calculating s are given in annex A.
item.
The “R” method (given in annex C) is simple to calculate, but
b) In terms of the knowledge gained, the advantage lies
with inspection by variables as the more precise information requires a somewhat larger sample size for the same AQL. Also
it has the undesirable property that for samples of size 10 or
obtained indicates how good the product is; earlier warning
will be given if the quality is slipping. more the acceptability of a lot can depend upon how the
sample is divided into sub-groups.
c) An attributes scheme can be more readily understood
and accepted; for example, it may at first be difficult to ac-
Initially, it will be necessary to begin with the “s” (or the
cept that, when inspecting by variables, a lot can be re-
“R” method), but if the quality is satisfactory, the standard
jected on measurements taken of a sample that does not
switching rules will permit the responsible authority to com-
contain any nonconforming items. (See the example
mence reduced inspection and use a smaller sample size.
in 14.6.)
The question then is, if the variability is under control and lots
d)
A comparison of the size of the samples required for the
continue to be accepted, will it be economical to change to the
same AQL from standard plans for inspection by attributes
“a” method ?
(i.e. from IS0 2859) and the standard plans in this Inter-
national Standard is given in table I-B. It will be seen that
The size of the sample will generally be smaller and the accept-
the smallest samples are required by the “a” method, used
ability criteria become simpler. (See clauses 15.2 and 15.3.) On the
when the standard deviation of the process is known.
other hand, it will still be necessary to calculate s for record pur-
poses and to keep the control charts up to date. (See clause 18.1
e) Inspection by variables is appropriate particularly in
conjunction with the use of control charts for variables.
f) Variables sampling has a substantial advantage when
11 Choice of inspection level and AQL
the inspection process is expensive, for example in the case
of destructive testing.
In standard sampling plans, the inspection level in conjunction
with the size of the lots and the AQL determines the size of the
g) A variables scheme becomes less suitable as the
sample to be taken, and governs the severity of the inspection.
number of measurements to be taken on one item in-
The appropriate OC curve given in one of the tables V-B to V-P
creases, as each characteristic has to be considered
shows the extent of the risk that is involved in such a plan.
separately. It may be advantageous to apply “attributes” to
the majority of the characteristics and “variables” to one or
The choice of the inspection level and AQL is governed by a
two of the more important requirements, for example proof
number of factors, but is mainly a balance between the total
load tests, safety and reliability requirements.
cost of inspection and the consequences of nonconforming
h) The use of this International Standard is only applicable items passing into service.
when there is reason to believe that the distribution of
measurements is normal. In case of any doubt, the respon-
The normal practice is to use inspection level II, unless special
sible authority should be consulted. circumstances indicate that another level is more appropriate.
IS0 3951 : 1989 (E)
12 Choice of sampling plan this International Standard may be selected by using
diagram A. The intersection of a vertical line through the
acceptable value for the limiting quality and a horizontal
12.1 Standard plans
line through the desired quality with a 95 % probability of
acceptance (approximately equal to AQL) will lie on,or under, a
The standard procedure can be used only when the production
sloping line indexed with the sample size code letter of a stan-
of lots is continuous.
dard plan which meets the specified requirements. This should
be verified by inspecting the OC curve given in table V for this
The standard procedure, with its semi-automatic steps from lot
code letter and AQL.
size to sample size, using inspection level II and beginning with
the “s” method, has been found in practice to produce
If the lines intersect at a point above the line marked P (see
workable sampling plans; but it assumes that the order of
diagram A), this implies that, for example, a sample of over 200
priority is first the AQL, second the sample size and last, the
would be necessary for the s method and the specification can-
limiting quality.
not be met by the plans in this International Standard.
The acceptability of this system is due to the fact that the con-
sumer is protected by the switching rules (see clause 191,
12.2 Special plans
which quickly increase the severity of inspection and finally ter-
minate it, if the quality of the process is worse than the AQL.
If standard plans are not acceptable, it will be necessary to
devise a special plan. The choice is then to decide which com-
NOTE - It should also be remembered that the limiting quality is the
bination of AQL, limiting quality, and sample size is most
quality which if offered for inspection would have a 10 % probability of
suitable, remembering that these are not independent, for,
acceptance. The actual risk taken by the consumer therefore also
depends on the probability of goods of this low quality being offered
when any two have been chosen, the third follows.
for inspection.
NOTE - This choice is not completely unfettered; the fact that the size
However, if, in certain circumstances, the limiting quality has a
of the sample is necessarily a whole number imposes some constraints.
higher priority than the sample size (for example, when only a
If a special scheme is necessary it should be devised only with the
limited number of lots are being produced), a suitable plan in assistance of a statistician experienced in quality control.
IS0 3951 :1989(E)
Section three: Operation of a variables sampling plan
Thus, if only the upper specification limit U is given, the lot is
13 Preliminary operations
Before starting inspection by variables, check
acceptable if 0” > k
a) that the distribution can be considered to be normal and
not acceptable if Q” < k
that production is considered to be continuous;
b) whether the “s” (or “R ”) method is to be used initially
Or, if only the lower specification limit L is given, the lot is
or whether the standard deviation is stable and known, in
which case the “a” method should be used;
acceptable if QL > k
c) that the inspection level to be used has been
not acceptable if QL < k
designated. If none has been given, inspection level II shall
be used;
When both U and L are given (k values are different if the AQLs
d) that the AQL has been designated and that it is one of
are different for the upper limit and the lower limit), the lot is
the preferred AQLs for use with this International Standard.
If it is not, the tables are not applicable;
acceptable if both 0, > kL and Q” > kU
e) if a double specification limit has to be met, whether the
not acceptable if either QL < kL or QU < k”
limits are separate or combined and, if the limits are
separate, whether AQLs are determined for each limit.
Example
The maximum temperature of operation for a certain device is
14 Standard procedure for “sl’ method
specified as 60 OC. Production is inspected in lots of 100 items.
Inspection level II, normal inspection with AQL = 2 ’5 % is to
14.1 Obtaining a plan
be used. From table I-A, the sample size code letter is F; from
table II-A it is seen that a sample size of 10 is required and that
The procedure for obtaining a plan is as follows :
the acceptability constant k is 1 ’41. Suppose the measure-
a) With the inspection level given (normally this will be II) ments are as follows: 53 OC; 57 “C; 49 OC; 58 “C; 59 “C; 54 OC;
and with the lot size, obtain the sample size code letter 58 OC; 56 OC; 55 OC; 50 OC. Compliance with the acceptability
using table I-A.
criterion is to be determined.
b) With this code letter and the AQL, enter table II-A and
obtain the sample size n and acceptability constant k.
Information needed Values obtained
c) Taking a random sample of this size, measure the
Sample size : n 10
characteristic x in each item and then calculate X, the
sample mean, and s, the estimated standard deviation (see
Sample mean X: C x/n
w9
annex A). lf X is outside the specification limit, the lot can
be judged unacceptable without calculating s. It may,
Standard deviation s : J C (xi - FP/(n - 1) 3,414
however, be necessary to calculate s for record purposes.
(See A.l.2, annex A.) ’
14.2 Acceptability criteria for single or separate
Specification limit (upper) : U 60
double specification limits
QU = (U - X)/s 1,494
If single or separate specification limits are given, calculate the
quality statistic
Acceptability constant : k (see table II-A) 1 ’41
u-x
QU= -
Acceptability criterion : compare 0, with k 1,494 > 1 ’41
S
and/or The lot meets the acceptability criterion, and is therefore accep-
table.
X-L
QL = -
S
14.3 Graphical method for a single specification
as appropriate,
limit
then compare the quality statistic (Q, and/or Q,) with the
When a graphical criterion is desired, draw the line
acceptability constant k obtained from table II-A for normal
inspection. If the appropriate quality statistic is greater than or
j+
U - k s (for an upper limit) or
equal to the acceptability constant, the lot is acceptable; if less,
it is not acceptable. X = L + k s (for a lower limit),
IS03951 :I989 (I3
as appropriate, on graph paper with Xas the vertical axis and s stants are k” = I,57 and kL = 254 respectively. Suppose the
as the horizontal axis. When the inspection concerns an upper sample delay times are as follows:
specification limit the accept zone is the zone below the line.
6,95 6,65
When a lower specification limit is considered, the accept zone 6 ’04
6 ’68 663
is the zone above the line. Using the values of s and X 6,34 6,15
6 ’40 6 ’4 604
7,15 6,70 6,51
calculated from the measurements obtained from a sample (see 644 659
7,17
annex A for the calculation of s), plot the point (s, XI on the 6,35 6,25 6,96
W33
graph. If this point lies in the accept zone, the lot is acceptable; 6,80 6,15 6,25 6,57
if outside, it is not acceptable. 6,52 6,59 6,86 6,57 6,91
6,29 6,67 6,67
663 6,70
Example
Compliance with the acceptability criteria is to be determined.
Using the data given in the example in 14.2, mark the point
u= 60 on the X (vertical) axis and draw a line through this
Information needed Value obtained
point with a slope - k [as k = 1,41, this means the line passes
through points (s = 1, X = 58,591, (s = 2, X = 57,18), (s = 3,
Sample size: n 35
X=
55,771, etc.]. Select a suitable point and draw a straight
line through it and (s = 0, X = 60), i.e. U. The accept zone is
Sample mean X: C x/n 6,55 s
then the area under this line. The calculated values of s and X
are 3,414 and 54,9. Plotting the point (s, X), it will be seen from
Sample standard deviation s :
figure 1 that it lies just inside the accept zone; the lot is accep-
- XP/(n - I)
-\/c (Xi 0,31 s
table.
(See A.l.2, annex A.)
The graph can be prepared before beginning the inspection of a
series of lots. Then, for each lot plot the point (s, X) and decide
Upper specification limit: U 9,0 s
whether or not the lot is acceptable.
QU = (U - X)/s
7 ’90
z U=60
t k =I,41
Acceptability constant: ku (see table II-A) I,57
Lower specification limit: L 4,0 s
-
QL = (x - L)ls 8,23
zone
56 -
Acceptability constant: kL (see table II-A)
2 ’54
Acceptability criterion: is QU > kU 7,90> I,57
and QL > kL?
and 8,23 > 254
The lot meets the acceptability criteria, and is acceptable.
t-lgure 1 - Example of the use of an acceptance chart
for a single speci f ication limit : 3” method
14.5 Graphical method for separate double
specification limits
14.4 Numerical method for single or separate
double specification limi
ts
When a graphical criterion is desi red for separate double
specification limits, draw the lines
Example
ji-=
U - k ”s (for the upper limit) and
A certain pyrotechnic delay has a specified minimum delay time
of 4,0 s and a maximum time of 9,0 s. Production is inspected
X = L + kL s (for the lower limit)
in lots of 1 000 items and inspection level II, normal inspection,
is to be used with an AQL of 0,l % applied to the lower limit on graph paper with Xas the vertical axis and s as the horizontal
and an AQL of 2,5 % applied to the upper limit. From table I-A axis. Using the values of s and x calculated from the
it is seen that the sample size code letter is J; from table I-B it is measurements obtained from a sample, plot the point (s, X) on
seen that the sample size is 35 for the “s” method and from the graph. If this point lies in the accept zone, the lot is accep-
table II-A it is found that the upper and lower acceptability con- table; if outside, it is not acceptable.
IS0 3951 :1989 (E)
and tightened inspection be copied (or traced) onto graph
Example
paper. (A sheet of the special graph paper needed for this is
given in annex D.) The scales should be adjusted so that s and
Using the data given in the example in 14.4 mark the point
X can be plotted directly (e.g. the upper limit is given
u= 9,O on the x (vertical) axis and draw a line through this
instead of I,0 and the lower limit instead of 0 on the STscale).
point with a slope - kU [as kU = 1 ’57, this means that the line
will pass through the point (s = 1, X = 7/W]. Also mark the
Then plot on the chart the values of s and xfound from the
point L = 4,O on the Taxis and draw a line through this point
sample and, if the point lies inside the accept zone, the lot is ac-
with slope + kL [as kL = 254, this means that the line will pass
ceptable; if outside, it is not acceptable.
through the point (s = 1, X = 6,54)]. The accept zone is then
the area bounded by these lines and the Kaxis. The calculated
-
NOTE - For sample size code letters B and C (i.e. sample sizes 3 and
values of s and x are 0 ’31 and 6 ’55. Plotting the point
4), the accept zone is bounded by four straight lines: the Taxis
...
NORME
ISO
INTERNATIONALE
l
1989-09-15
Règles et tables d’échantillonnage pour les
contrôles par mesures des pourcentages de non
conformes
Sampling procedures and charts for inspection by variables for percent
noncon forming
ISO3951:1989 (FI
Avant-propos
L’ISO (Organisation internationale de normalisation) est une fédération mondiale
d’organismes nationaux de normalisation (comités membres de I’ISO). L’élaboration
des Normes internationales est en général confiée aux comités techniques de I’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 I’ISO participent également aux travaux. L’ISO col-
labore étroitement avec la Commission électrotechnique internationale (CEI) en ce qui
concerne la normalisation électrotechnique.
Les projets de Normes internationales adoptés par les comités techniques sont soumis
aux comités membres pour approbation, avant leur acceptation comme Normes inter-
nationales par le Conseil de I’ISO. Les Normes internationales sont approuvées confor-
mément aux procédures de I’ISO qui requièrent l’approbation de 75 % au moins des
comités membres votants.
La Norme internationale ISO 3951 a été élaborée par le comité technique ISO/TC 69,
Applications des méthodes statistiques.
Cette deuxième édition annule et remplace la première édition (ISO 3951 : 1981) dont
elle constitue une révision technique.
Les principaux changements par rapport à la première édition sont les suivants:
a) une distinction a été faite entre l’écart-type maximal du processus (ETMP)
selon la méthode w)) et l’écart-type maximal de l’échantillon (ETME) selon la
méthode «s»;
b) toutes les courbes d’acceptation pour la méthode WJD ont été tronquées aux
valeurs appropriées de l’écart-type maximal du processus;
c) la terminologie a été alignée sur celle de I’ISO 2859 et I’ISO 3534.
L’attention des utilisateurs est attirée sur le fait que toutes les Normes internationales
sont de temps en temps soumises à révision et que toute référence faite à une autre
Norme internationale dans le présent document implique qu’il s’agit, sauf indication
contraire, de la derniére édition.
0 ISO 1989
Droits de reproduction réservés. 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’éditeur.
Organisation internationale de normalisation
Case postale 56 l CH-1211 Genève 20 l Suisse
.
Imprimé en Suisse
ii
ISO 3951 : 1989 (F)
Sommaire
Page
Section un : Gén&alités. . 1
1 Objet et domaine d’application .
2 Références .
3 Définitions et symboles .
4 Niveau de qualité acceptable (NQA) .
......... 4
5 Règles pour la modification du contrôle normal, renforcé et réduit
6 Relations avec I’ISO 2859. .
7 Production non continue et courbes d’efficacité . 5
Section deux : Choix d’un plan d’échantillonnage .
8 Généralités sur le choix d’un plan .
9 Choix entre mesures et attributs .
10 Choix d’une méthode .
11 Choix du niveau de contrôle et du NQA .
12 Choix d’un plan d’échantillonnage .
Section trois : Mise en œuvre d’un plan d’échantillonnage par mesures . . 8
13 Opérations préliminaires indispensables .
............... 8
14 Procédure normalisée pour l’application de la méthode WH
Procédure normalisée pour l’application de la méthode ((ct)) . 13
Déroulement d’un contrôle continu . 16
17 Normalité et aberrants . 16
18 Enregistrement des résultats . 16
............... 17
19 Mise en œuvre des règles pour la modification du contrôle.
20 Suspension et reprise du contrôle. . 17
21 Courbes d’acceptation dans le cas du contrôle renforcé et du contrôle réduit
22 Passage à la méthode ((0)) .
. . .
III
Section quatre : Tables et graphiques . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
l
Tables
Lettres-code d’effectif d’échantillon et niveaux de contrôle . . . 19
I-A
Lettres-code d’effectif d’échantillon et effectif
I-B
d’échantillon dans un contrôle normal . . . . . . . . . . . . . . . . . . . . . 20
II-A Plans d’échantillonnage simples pour le contrôle normal
(table générale) : méthode GSD . . . . . . . . . . . . . . . . . . . . . . . . . . .
II-B Plans d’échantillonnage simples pour le contrôle renforcé
(table générale) : méthode NS)) . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
II-C Plans d’échantillonnage simples pour le contrôle réduit
(table générale) : méthode NS)) . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
Ill-A Plans d’échantillonnage simples pour le contrôle normal
(table générale) : méthode w)) . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
Ill-B Plans d’échantillonnage simples pour le contrôle renforcé
(table générale) : méthode ((0)) . . . . . . . . . . . . . . . . . . . . . . . . . . .
III-C Plans d’échantillonnage simples pour le contrôle réduit
(table générale) : méthode WN . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
Valeurs de fs pour l’écart-type maximal de l’échantillon
IV-S
(ETME) : méthode w) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
IV-a Valeurs de fa pour l’écart-type maximal du processus
(ETMP) : méthode ((0)) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
\
Tables et courbes d’efficacité des plans d’échantillonnage
V (V-B à V-P)
simples (lettres-code d’effectif d’échantillon B à P) . . . . . . . . . . 28
VI-A Correspondance entre le plan de contrôle renforcé
et le plan de contrôle normal . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
Correspondance entre le plan de contrôle réduit
VI-B
et le plan de contrôle normal . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
Graphiques
A Lettres-code d’effectifs d’échantillon des plans d’échantillonnage
normalisés tels que, pour des qualités spécifiées, les
probabilités d’acceptation soient de 95 % et 10 % . . . . . . . . . . . 43
s-D à s-P Courbes d’acceptation pour des limites de spécification
doubles combinees : méthode NS)) . . . . . . . . . . . . . . . . . . . . . . . . 44
a-c à a-P Courbes d’acceptation pour des limites de spécification
doubles combinées : méthode ((0)). . . . . . . . . . . . . . . . . . . . . . . . 56
Annexes
A Calculdesetdea. . 69
B Théorie statistique . 72
C Plans d’échantillonnage pour la méthode «R)). . 83
................................. 106
D Papier millimétré pour la méthode NS))
Bibliographie .
iv
NORME INTERNATIONALE
ISO 3951 : 1989 (FI
Règles et tables d’échantillonnage pour les contrôles
par mesures des pourcentages de non conformes
Section un : Généralités
1 Objet et domaine d’application d) lorsqu’un contrat ou une norme définit une limite
supérieure de spécification L,, une limite inférieure de
spécification Lit ou les deux; un produit est qualifié de non
1.1 Objet
conforme si la mesure de son caractère de qualité x satisfait
à l’une des inégalités suivantes:
1.1.1 La présente Norme internationale présente les plans et
x > L, . . .
(1)
règles d’échantillonnage pour les contrôles par mesures. Elle
complète I’ISO 2859. Les cahiers des charges, les contrats, les
X < Li . . .
(2)
instructions de contrôle ou autres textes peuvent se référer à la
soit x > L,, soitx < Li
. . . (3)
présente Norme internationale et à I’ISO 2859 lorsque l’autorité
responsable le prescrit, et les dispositions qu’elles contiennent
Les inégalités (1) et (2) répondent aux cas d’une limite unique
doivent être respectées. L’autorité responsable doit être dési-
de spécification et (3) au cas d’une limite double de spkifi-
gnée dans l’un de ces textes.
cation. Dans cette dernière situation, une distinction supplé-
mentaire est apportée entre les limites doubles séparées ou
1.1.2 Le but des méthodes présentées dans la présente
combinées, selon que le NQA est appliqué à chaque limite
Norme internationale est d’assurer que les lots de qualité accep-
séparément ou aux deux limites combinées (voir chapitre 4).
table ont une forte probabilité d’être acceptés et que la probabi-
lité de ne pas accepter les lots de qualité inférieure est aussi éle-
vée que possible.
2 Références
1.1.3 En accord avec I’ISO 2859, le pourcentage de produits ISO 2854, Interprétation statistique des données - Techniques
d’estimation et tests portant sur des moyennes et des
non conformes dans les lots est utilisé pour définir la qualité de
variances.
ces lots et des procédés de fabrication en question.
ISO 2859, Règles et tables d’échantillonnage pour les contrôles
1.2 Domaine d’application
par attributs,
La présente Norme internationale est principalement destinée à
ISO 3534, Statistique - Vocabulaire et symboles.
être utilisée dans les conditions suivantes :
ISO 5479, Tests de normalité. 1)
a) lorsque les règles de contrôle doivent être appliquées à des
séries continues de lots de produits individualises tous
I S 0 5725, Fidélité des me thodes d’essai - Dé termina tion de la
fournis par un seul producteur utilisant un seul procédé de
répé tabilité et de la reproductibilité d’une méthode d’essai nor-
fabrication. S’il y a différents producteurs, la présente Norme
malisée par essais in terlabora toires.
internationale doit être appliquée à chacun d’eux séparément;
b) lorsqu’un seul caractère de qualité, x, de ces pro-
duits est pris en considération, il doit être mesurable sur
3 Définitions et symboles
une échelle continue. Si plusieurs caractéres du produit
sont importants, la présente Norme internationale doit être
3.1 Définitions
appliquée à chacun d’eux séparément;
c) lorsque la fabrication est stable (sous contrôle stasti- Pour les besoins de la présente Norme internationale, les défini-
que) et que le caractère de qualité x est distribué suivant une tions données dans I’ISO 3534 et I’ISO 2859 ainsi que les défini- .
loi normale ou voisine d’une loi normale;
tions suivantes s’appliquent.
1) Actuellement au stade de projet.
ISO 3951 : 1989 (FI
3.1.1 contrôle par mesures (inspection by variables) : Con- 3.1.10 limite de spécif ication (specification limit) : Valeur
trole dans lequel on mesure un caractére quantitatif lie à chacun limite (minimale ou maximale) spécifiée pour un caractere
des individus d’une population ou d’un échantillon prélevé dans quantitatif.
cette population.
3.1.11 limite inférieure de spécification (LJ [lower specifi-
3.1.2 echantillonnage par mesures en vue d’acceptation
cation limit (L)] : Limite de spécification qui définit la limite infé-
(acceptance sampling by variables) : Procédure selon laquelle
rieure de reconnaissance de conformité d’une unité individuelle
un caractère spécifié est mesuré pour établir statistiquement
dans une fabrication ou un service.
I’acceptabilité du lot à partir des résultats obtenus sur les indivi-
dus d’un échantillon.
3.1.12 limite supérieure de spécification (L,) [Upper speci-
fication limit (U)] : Limite de spécification qui définit la limite
3.1.3 niveau de qualité acceptable (NQA) [acceptable
supérieure de reconnaissance de conformité d’une unité indivi-
quality level (AQL)] : Dans le cas d’une série continue de lots,
duelle dans une fabrication ou un service.
niveau de qualité qui, pour le contrôle par échantillonnage, est
la limite du pourcentage satisfaisant de non conformes de la
caractéristique moyenne de la qualité de production. (Voir cha-
3.1.13 limite unique de spécification (single specification
pitre 4.)
limit) : Terme utilisé lorsqu’il n’est spécifié qu’une seule limite.
3.1.4 qualite limite (limiting quality) : Pour un lot considéré
3.1.14 limites de specification doubles separées (sepa-
isolément, niveau de qualité qui, dans un contrôle par échantil-
rate double specification limits) : Terme utilisé lorsque sont
lonnage, correspond à une faible probabilité d’acceptation
spécifiées à la fois une limite supérieure et une limite inférieure
(dans la présente Norme internationale: 10 %). (Voir 12.1.)
et quand des NQA distincts sont appliqués à chacune des limi-
tes individuellement. (Voir 4.3.)
3.1.5 non-conformité (nonconformity) : Défaut de carac-
tére de qualité d’un produit ou service à ne pouvoir répondre à
3.1.15 limite de spécification double combinée (combi-
une exigence spécifiée, et dont l’évaluation ne dépend pas
ned double specification limit) : Terme utilisé lorsque sont spé-
essentiellement de la durée de temps.
cifiées à la fois la limite supérieure et la limite inférieure et quand
un NQA donné s’applique à l’ensemble des deux limites et cor-
Selon leur degré d’importance, les non-conformités sont géné-
respond au pourcentage total de non conformes. (Voir 4.3.)
ralement classées comme suit:
ayant une incidence grave
Classe A. Type de non-conformités
3.1.16 constante d’acceptabilité (k) (acceptability cons-
sur le produit ou le service. Dans l’échantillonnage en vue
tant) : Constante dépendant de la valeur assignée au niveau de
d’acceptation, on fixera pour de tels types de non-conformités
qualité acceptable et de l’effectif de l’échantillon. (Voir 14.2 et
une valeur très faible pour le NQA.
15.2, éventuellement chapitre C.5 de l’annexe C.)
Classe B. Type de non-conformités ayant une incidence moins
3.1.17 statistique de qualité (0) (quality statistic) : Fonc-
importante et pour lesquelles sera fixée une valeur du NQA
supérieure à celles de la classe A et inférieure à celles de la tion de la limite de spécification, de la moyenne de l’échantillon
et de l’écart-type de l’échantillon. La décision sur le lot est prise
classe C, si cette dernière existe, et ainsi de suite.
en comparant 0 avec la constante d’acceptabilité k. (Voir 14.2
Le nombre des classes et l’affectation d’une non-conformité et 15.2, éventuellement chapitre C.5 de l’annexe C.)
dans une classe donnée doivent être adaptés aux exigences de
qualité de la situation particulière considérée.
3.1.18 statistique de qualité correspondant à la limite
inférieure (Qi) [lower quality statistic (QL)l : Fonction de la
3.1.6 unité non conforme (nonconforming unit) : Unité de
limite inférieure de spécification, de la moyenne de l’échantillon
produit ou de service contenant au moins une non-conformité. et de l’écart-type de l’échantillon. La décision sur le lot est prise
en comparant Qi avec la constante d’acceptabilité k. (Voir 14.2
et 15.2 ou chapitre C.5 en annexe C).
3.1.7 méthode (w) (‘Y method) : Méthode permettant de
décider si un lot est acceptable à partir de l’écart-type de
l’échantillon. (Voir chapitre 14.)
3.1.19 statistique de qualité correspondant à la limite
supérieure (Q,) [Upper quality statistic (QU)I : Fonction de la
limite supérieure de spécification, de la moyenne de I’échantil-
3.1.8 méthode (WH (“0” méthod) : Méthode permettant de
Ion et de l’écart-type de l’échantillon. La décision sur le lot est
décider si un lot est acceptable à partir de la connaissance que
prise en comparant Q, avec la constante d’acceptabilité k.
l’on a de l’écart-type du processus. (Voir chapitre 14.)
(Voir 14.2 et 15.2 ou chapitre C.5 en annexe C.)
3.1.9 methode c(R)) (‘Y?” method) : Méthode permettant de
décider si un lot est acceptable en adoptant une estimation indi-
3.120 écart-type maximal de l’échantillon (ETME) [maxi-
recte de l’écart-type du processus à partir de l’étendue
mum sample standard deviation (MSSD)] : Dans des conditions
moyenne des mesures effectuées par sous-groupes d’individus
données, le plus grand écart-type de l’échantillon acceptable.
d’un échantillon. (Voir annexe C.)
(Voir 14.6 et B.8.3 dans l’annexe B.)
ISO 3951 : 1969 (FI
3.1.21 écart-type maximal du processus (ETMP) [maxi- & Valeur d’acceptation maximale.
mum process standard deviation (MPSD)]: Dans des condi-
tions données, le plus grand écart-type du processus accepta- p Moyenne du processus
ble. (Voir 15.3 et B.5.2 dans l’annexe B.)
ct Écart-type du processus (a*, carré de l’écart-type, est
appelé variante).
3.122 règles de modification du contrôle (switching
rules) : Instructions, contenues dans une procédure d’échantil-
C «Somme de)) (par exemple, C x = somme de tous
lonnage, passage d’un plan d’échantillonnage à un autre, à par-
les x).
tir d’un compte-rendu prouvé de la qualité. (Voir chapitre 19.)
n
Somme de tous les Xi quand i prend toutes les
xi
c
3.2 Symboles valeurs de 1 à n.
i= 1
«Supérieur à» (par exemple, a > b signifie que a est
Les symboles employés sont les suivants:
supérieur à b).
& Facteur, donné dans la table IV-s, qui relie l’écart-type
«Supérieur ou égal à» (par exemple, a > b signifie que
maximal de l’échantillon à la différence entre L, et Li.
a est supérieur ou égal à 6).
fa Facteur, donné dans la table IV-a, qui relie l’écart-type
«Inférieur à» (par exemple, a < b signifie que a est
maximal du processus à la différence entre L, et Lia
inférieur à 6).
k Constante d’acceptabilité quand on utilise la méthode
«Inférieur ou égal à» (par exemple, a < b signifie que a
as)), mn) ou (db.
est inférieur ou égal à b).
K Constante d’acceptabilité lorsque ,u et 0 sont connus.
3.3 Bibliographie
Li Limite inférieure de spécification («i», en indice à une
Une bibliographie des documents utilisés pour l’élaboration de
variable, signifie valeur de cette variable à Li).
la présente Norme internationale est donnée dans l’annexe D.
L, Limite supérieure de spécification ((w, en indice à une
variable, signifie valeur de cette variable à L,).
4 Niveau de qualité acceptable (NQA)
n Effectif d’échantillon (nombre d’unités dans un échan-
4.1 Définition
tillon).
Dans le cas d’une série continue de lots, le NQA est le niveau de
N Effectif du lot (nombre d’unités dans un lot).
qualité qui, pour le contrôle par échantillonnage, est la limite du
pourcentage satisfaisant de non conformes de la caractéristi-
pa Probabilité d’acceptation.
que moyenne de la qualité de production.
0 Statistique de qualité.
4.2 Utilisation
Qi Statistique de qualité correspondant à la limite infé-
rieure.
Le NQA est utilisé en même temps que la lettre-code d’effectif
d’échantillon pour repérer les plans d’échantillonnage décrits
Q, Statistique de qualité correspondant à la limite supé-
dans la présente Norme internationale.
rieure.
4.3 Spécification du NQA
s Écart-type d’un échantillon (estimation de l’écart-type du
processus)
Le (les) NQA utilisé(s) doivent être fixé(s) dans le contrat de
spécification du produit ou par l’autorité responsable. Si des
n
limites de spécification supérieure et inférieure sont données,
(Xi - X)*
des NQA séparés peuvent être attribués aux limites individuel-
lz
i= 1 les, qui sont alors considérées comme des limites de spécifica-
=
S
tion doubles séparées. Alternativement, un seul NQA peut être
n-l
li
spécifié pour le pourcentage total de non conformes s’appli-
quant à la fois à la limite supérieure et à la limite inférieure; on a
(Voir aussi annexe A.)
alors affaire à une limite de spécification double combinée.
x Valeur mesurée d’un caractère dans l’échantillon.
4.4 NQA recommandés
X Moyenne arithmétique de x pour un échantillon de n
individus.
Onze NQA donnés dans la présente Norme internationale, dont
les valeurs s’échelonnent de 0,lO % à 10 % de non conformes,
-
Xi Valeur d’acceptation minimale. sont dits «NQA recommandés». Si, pour n’importe quel produit
60 3951 : 1989 (FI
5.6 Des détails relatifs aux règles de modification du contrôle
ou service, on spécifie une valeur de NQA autre qu’une autre
sont donnés dans les chapitres 19 et 20.
valeur recommandée, la présente Norme internationale n’est
alors pas applicable. (Voir 12.2.)
Deux NQA supplémentaires, 0,085 % et 15 %, sont donnés
6 Relations avec I’ISO 2859
dans le but de compléter la gamme des plans que nécessite
l’application des règles de modification du contrôle. (Voir cha-
pitres 19 et 21.) Les plans et les courbes désignés par un NQA 6.1 Similitudes
de 0,065 % ou de 15 % ne,seront utilisés que lorsque le NQA,
dans le contrôle normal, est de 0,lO % ou de 10 % respective-
a) La présente Norme internationale complète I’ISO 2859;
ment, et que les règles de modification du contrôle sont invo-
les deux documents ont une philosophie commune et,
quées.
autant que possible, leurs règles et vocabulaire sont identi-
ques.
4.5 Remarque
b) Tous deux utilisent le NQA pour indexer les plans
d’échantillonnage, et les valeurs recommandées utilisées
D’après la définition du NQA donnée en 4.1, il s’ensuit que la
dans ce document sont identiques à celles qui sont données
protection désirée ne peut être obtenue que si une série conti-
dans I’ISO 2859 pour la même étendue de valeurs (c’est-à-
nue de lots est présentée au contrôle.
dire de 0,l % à 10 %).
c) Dans les deux Normes internationales, l’effectif du lot et
4.6 Limitation
le niveau de contrôle (le niveau II étant recommandé à
défaut d’autres instructions) déterminent une lettre-code.
La fixation d’un NQA n’implique pas pour le fournisseur le droit
Des tables générales donnent ensuite, en fonction de la
de livrer sciemment des individus non conformes.
lettre-code et du NQA fixé, l’effectif de l’échantillon à préle-
ver et le critère d’acceptabilité suivant la méthode adoptée
(w), GO)), ou éventuellement, «RH). Des tables distinctes
5 Règles pour la modification du contrôle
sont données pour le contrôle normal, le contrôle renforcé
et le contrôle réduit.
normal, renforcé et réduit
d) Les règles pour la modification du contrôle sont essen-
5.1 Afin de décourager le producteur d’opérer avec un pour-
tiellement les mêmes.
centage de non conformes qui dépasse celui préconisé par le
e) La classification des non-conformités en fonction du
NQA, la présente Norme internationale prescrit de passer au
degré de leur importance en classe A, classe B, etc. n’est
contrôle renforcé dans le cas où les résultats du contrôle font
pas modifiée.
présumer que la qualité moyenne de production est inférieure à
celle qui est caractérisée par le NQA, et de suspendre entière-
ment le contrôle par échantillonnage dans le cas où le contrôle
6.2 Différences
renforcé n’encourage pas à temps le producteur à améliorer son
processus de fabrication.
a) Détermination de I’acceptabilité. L’acceptabilité,
dans un plan d’échantillonnage par attributs de I’ISO 2859,
5.2 Le contrôle renforcé et la règle d’interruption font partie est déterminée par le nombre d’unités non conformes trou-
intégrante de la présente Norme internationale et, par consé- vées dans l’échantillon; le critère d’acceptabilité dans le con-
quent, constituent des procédures d’application obligatoire
trôle par mesures est basé sur les estimations de la tendance
pour maintenir la protection obtenue pour un NQA donné. centrale et de la variabilité de la distribution des mesures
dans le lot, en relation avec les limites de spécification. Ces
notions sont traduites en termes de moyenne et d’écart-
5.3 La présente Norme internationale prévoit aussi la possibi-
type. Dans la présente Norme internationale, deux métho-
lité de passer à un contrôle réduit si les résultats du contrôle
la méthode w) que l’on applique
des sont présentées :
montrent que le pourcentage des non conformes dans la qua-
dans le cas où l’écart-type du processus 0 est inconnu et la
lité moyenne de fabrication est stable et meilleure que le NQA.
méthode ((0)) que l’on applique dans le cas où 0 est consi-
Cette méthode est cependant facultative (a la discrétion de
déré comme étant connu. Une troisième méthode, appelée
l’autorité responsable) l
méthode c(R)), est donnée dans l’annexe C. Dans le cas
d’une limite unique de spécification ou de deux limites sépa-
rées, l’acceptabilité peut être calculée à partir d’une formule
5.4 S’il existe des preuves suffisantes d’après les cartes de
(voir 14.2 et 15.21, mais elle est établie plus facilement par
contrôle (voir 18.1) que la variabilité est sous contrôle, il peut
une méthode graphique (voir 14.3). Dans le cas d’une limite
être envisagé de passer à la méthode ((a)). Si cette méthode est
double combinée, la présente Norme internationale prévoit
effectivement avantageuse, la valeur à adopter pour a est la
une méthode graphique (voir 14.4 et 15.3).
valeur stable de s.
b) Normalité. Dans I’ISO 2859, aucune exigence relative
5.5 S’il a été nécessaire d’interrompre le contrôle par échan- à la distribution des caractères n’est stipulée; par contre
dans la présente Norme internationale, il est nécessaire,
tillonnage, le contrôle renforcé ne peut pas être repris avant
que des actions aient été prises par le fabricant pour améliorer pour l’application correcte du plan, que les mesures soient
distribuées suivant une loi normale ou voisine de la normale.
la qualité du produit.
1s0 3951 : 1989 W-1
c) Courbes d’efficacité (OC curves). Bien que l’on
7 Production non continue et courbes
puisse concevoir un plan d’échantillonnage par mesures
d’efficacité
dont la courbe d’efficacité corresponde étroitement à celle
d’un plan par attributs, il ne serait pas possible d’établir,
7.1 Production non continue
dans la présente Norme internationale, toutes les courbes
d’efficacité correspondant à celles de I’ISO 2859 (indexées
a) Les méthodes d’échantillonnage contenues dans la pré-
selon la même lettre-code et le même NQA) sans que
sente Norme internationale ne sont pas destinées à être
l’effectif d’échantillon n’augmente avec le NQA pour une
appliquées dans des conditions différentes de celles qui sont
lettre-code d’effectif d’échantillon donnée. Pour la méthode
spécifiées en 1.2, par exemple pour un lot isolé ou un nom-
(~1, on a gardé le même effectif d’échantillon pour un effec-
bre limité de lots, pour lesquels le contrôle renforcé et les
tif de lot donné, dans toute la gamme des NQA; pour la
règles d’interruption ne peuvent pas être appliqués.
méthode w)), cette contrainte a été levée, afin d’associer
b) Dans ces conditions, le concept de NQA n’est plus perti-
aussi étroitement que possible la courbe d’efficacité de la
nent, dans la mesure où le consommateur ne se soucie que de
méthode w) et celle de la méthode HO)), à la fois au NQA et
la qualité du nombre limité des lots présentés, cessant d’exer-
à la qualité limite.
cer des contrôles sur la qualité du processus de fabrication
Le NQA indique cependant une qualité présentée qui a une
d) Probabilité d’acceptation pour un NQA. La proba-
grande probabilité d’acceptation et peut donc servir de repé-
bilité qu’un lot, dont la qualité est égale à celle qui est carac-
rage d’un plan d’échantillonnage.
térisée par le NQA, soit accepté augmente avec l’effectif de
l’échantillon et suit une loi analogue mais non identique à
7.2 Courbes d’efficacité
celle qui est utilisée dans I’ISO 2859.
a) Le degré de protection du client obtenu pour chacun
des différents plans d’échantillonnage de cette norme peut
e) Effectifs d’échantillon. Les effectifs d’échantillon
néanmoins être évalué à partir des courbes d’efficacité don-
pour le contrôle par mesures correspondant à des lettres-
nées sur les graphiques V-B à V-P et dans les tables V-B-l à
code données sont généralement plus petits que les effec-
V-P-l, qui doivent être consultés pour le choix d’un plan
tifs d’échantillon pour le contrôle par attributs correspon-
d’échantillonnage.
dant aux mêmes lettres-code.
b) Ces courbes sont destinées au contrôle normal pour
l’application de la méthode w) avec une seule limite de spé-
f) Plans d’échantillonnage doubles. Aucun plan
cification, mais elles fournissent une bonne approximation
d’échantillonnage double n’est donné dans la présente
dans le cas d’une limite de spécification double combinée.
Norme internationale.
c) Ces courbes fournissent également une bonne approxi-
mation des courbes d’efficacité pour les plans d’échantillon-
g) Limite de la qualité moyenne après contrôle
nage de la méthode ((0)) (et de la méthode ((R))I repérés par
(AOQL). Lors d’essai destructifs ou coûteux, dans lesquels
la même lettre-code et le même NQA, sauf si l’effectif de
le contrôle et le remplacement à 100 % des lots non accep-
l’échantillon est petit.
tés n’est pas possible, le concept d’AOQL ne peut pas être
appliqué. Dans la mesure où des plans de contrôle par d) Les courbes d’efficacité séparées ne sont pas présen-
mesures sont généralement appliqués dans de tels cas, les tées pour le contrôle renforcé ou réduit, mais les courbes
tables correspondantes d’AOQL ne sont pas présentées peuvent être trouvées parmi celles données pour le contrôle
dans la présente Norme internationale. normal (voir tables VI-A et VI-B).
ISO 3951 : 1989 (FI
.
Section deux: Choix d’un plan d’échantillonnage
la distribution des mesures est normale. S’il existe un doute,
8 Généralités sur le choix d’un plan
l’autorité responsable doit être consultée.
Le choix du plan par mesures le mieux approprié, s’il en existe
un, requiert de l’expérience, du jugement et une assez bonne
NOTES
connaissance a la fois de la statistique et du produit à contrôler.
1 Des tests de normalité figurent dans la section deux de I’ISO 2854
La présente section vise à suggérer, aux responsables chargés
qui donne des exemples de méthodes graphiques pouvant être utilisées
de déterminer les plans d’échantillonnage, les considérations
pour vérifier qu’une distribution de données est suffisamment proche
qu’ils doivent avoir présentes à l’esprit quand ils décident si un
de la normale pour justifier l’utilisation de plans d’échantillonnage par
plan par mesures conviendrait bien et quels sont les choix à
mesures.
faire pour sélectionner un plan normalisé approprié.
2 Une documentation plus précise concernant les tests de normalité
est donnée dans I’ISO 5479.
9 Choix entre mesures et attributs
10 Choix d’une méthode
La première question qui se pose est de décider s’il est préféra-
ble de faire un contrôle par mesures plutôt qu’un contrôle par
Si l’on a décidé d’appliquer un contrôle par mesures, la pre-
attributs. II est recommandé de prendre en considération les
mière question qui se pose est de savoir quelle est la méthode à
points suivants :
employer : la méthode w) ou la méthode ((CT)) (éventuellement,
la méthode «R»).
a) Au point de vue économique, on doit comparer le coût
total d’un contrôle relativement simple d’un plus grand
La méthode ((0)) est la plus économique au point de vue de
nombre d’individus à l’aide d’une méthode par attributs à
l’effectif de l’échantillon mais, avant que cette méthode puisse
celui d’une procédure généralement plus élaborée, exigée
être employée, la valeur de ~0)) doit être déterminée.
par une méthode par mesures qui revient généralement plus
cher en temps et en argent dépensés par élément contrôlé.
Si l’on considère l’effectif d’échantillon, la méthode w) pré-
b) Au point de vue de l’information obtenue, l’avantage
sente un léger avantage sur la méthode «R» mais la détermina-
revient au contrôle par mesures pour lequel on obtient un
tion de s entraîne plus de calculs; l’importance et la difficulté de
renseignement plus précis sur le niveau de qualité du pro-
ceux-ci est plus apparente que réelle, particulièrement si l’on
duit : on sera prévenu plus tôt si la qualité dérive.
dispose d’un calculateur électronique. Les méthodes de calcul
de s sont données dans l’annexe A.
c) Une méthode par attributs peut être plus aisément com-
prise et acceptée; par exemple, il peut être difficile d’admet-
La méthode «R)) (donnée dans l’annexe C) est simple à calcu-
tre d’emblée que, lorsqu’on effectue un contrôle par mesu-
ler, mais elle exige un effectif d’échantillon un peu plus impor-
res, un lot puisse être rejeté à partir de mesures faites sur un
tant pour le même NQA. En outre, pour des échantillons
échantillon ne contenant pas de non conformes. (Voir
l’exemple en 14.6.) d’effectif 10 ou plus, I’acceptabilité d’un lot peut dépendre de la
manière dont est divisé l’échantillon en sous-groupes, ce qui
d) Une comparaison entre les effectifs d’échantillon exigés
constitue une propriété indésirable de la méthode.
pour obtenir le même NQA, en utilisant des plans normali-
sés du contrôle par attributs (c’est-à-dire I’ISO 2859) et les
Initialement, on devra commencer par la méthode w) (ou,
plans normalisés contenus dans la présente Norme interna-
éventuellement, la méthode c
tionale figure dans la table I-B. On remarquera que les
sante, les règles normalisées pour la modification du contrôle
échantillons les plus petits sont obtenus par la méthode
permettront, à l’autorité responsable, d’introduire un contrôle
w)), utilisée quand l’écart-type du processus est connu.
réduit et de prendre un effectif d’échantillon plus petit.
e) Le contrôle par mesures convient particulièrement bien
quand il est conjugué avec l’utilisation de cartes de contrôle
La question devient alors : si la variabilité est sous contrôle et si
par mesures.
les lots continuent d’être acceptés, serait-il plus économique de
passer à la méthode «a»?
f) L’échantillonnage par mesures offre un avantage consi-
dérable quand la méthode de contrôle est onéreuse, par
L’effectif de l’échantillon sera généralement moindre et le cri-
exemple, dans le cas d’un essai destructif.
tère d’acceptabilité sera plus simple. (Voir 15.2 et 15.3.) D’autre
g) Une méthode par mesures convient moins bien lorsque
part, il restera nécessaire de calculer s afin de pouvoir en enre-
s’accroît le nombre de caractères mesurés sur un même
gistrer les résultats et de tenir à jour les cartes de contrôle. (Voir
individu, puisqu’il y a lieu d’examiner chaque caractère
chapitre 18. )
séparément. On peut avoir avantage à appliquer le contrôle
par attributs à la majorité des caractères et le contrôle par
mesures à une ou deux des exigences les plus importantes,
11 Choix du niveau de contrôle et du NQA
par exemple : les essais de résistance à la charge, les exi-
gences de sécurité et de fiabilité.
Dans les plans d’échantillonnage normalisés, le niveau de con-
h) L’emploi de la présente Norme internationale n’est trôle, conjugué avec l’effectif des lots et avec le NQA, déter-
applicable que lorsqu’on a de bonnes raisons de croire que
mine l’effectif d’échantillon à prélever et régit la sévérité du
ISO 3951 : 1989 (F)
contrôle. La courbe d’efficacité appropriée, donnée dans l’une également de la probabilité que des marchandises de qualité inférieure
soient préSentées au contrôle.
des tables V-B à V-P, montre l’importance du risque inhérent
au ptan.
Cependant, si dans certaines circonstances, la qualité limite a
Le choix du niveau de contrôle et du NQA est fonction de plu- une priorité plus grande que l’effectif d’échantillon (par exem-
ple, quand la production comporte seulement un nombre limité
sieurs facteurs mais dépend principalement d’un compromis
des lots), il est possible de choisir un plan de la présente Norme
équilibré entre le coût total du contrôle et les conséquences de
la mise en service de produits non conformes. internationale bien adapté, en employant le graphique A.
L’intersection de la ligne verticale passant par la valeur choisie
La procédure habituelle consiste à appliquer un contrôle au de la qualité limite et de la ligne horizontale passant par la qua-
niveau II jusqu’à ce que des circonstances spéciales indiquent lité acceptable pour laquelle la probabilité d’acceptation est de
qu’un autre niveau est préférable. 95 % (approximativement égale au NQA) se situera sur ou
au-dessous d’une ligne inclinée indexée par la lettre-code
d’effectif d’échantillon du plan normalisé répondant aux condi-
12 Choix d’un plan d’échantillonnage tions spécifiées. Cela peut être vérifié en examinant la courbe
d’efficacité donnée dans la table V pour cette lettre-code
d’effectif d’échantillon et ce NQA.
12.1 Plans normalisés
Si ces lignes se coupent en un point situé au-dessus de la ligne
La procédure normalisée ne peut être utilisée que lorsque la
marquée «PH (voir graphique A), cela implique que, par exem-
production des lots est continue.
ple, un échantillon de plus de 200 individus serait nécessaire
pour la méthode s, et donc que les conditions spécifiées ne
Cette procédure, avec ses étapes semi-automatiques depuis
peuvent pas être remplies au moyen des plans normalisés.
l’effectif du lot jusqu’à l’effectif d’échantillon, en utilisant le
niveau de contrôle II et en commentant par la méthode (w),
s’est révélée, en pratique, fournir des plans d’échantillonnage
12.2 Plans spéciaux
fonctionnant de facon acceptable, mais elle suppose que
l’ordre de priorité soit d’abord le NQA, ensuite l’effectif
Si les plans normalisés ne peuvent pas être retenus, il est
d’échantillon et en dernier lieu la qualité limite.
nécessaire de concevoir un plan spécial. Le choix consiste alors
à décider quelle combinaison de NQA, qualité limite et effectif
La valeur de ce système est due au fait que le client est protégé
d’échantillon, convient le mieux, en se rappelant que ces élé-
par les règles de passage d’un niveau de contrôle à un autre
ments ne sont pas indépendants car il suffit d’en choisir deux
(voir chapitre 19) qui augmentent rapidement la sévérité du
pour que le troisième en découle.
contrôle et finalement y mettent fin, si la qualité de la fabrica-
tion est plus mauvaise que celle caractérisée par le NQA.
NOTE - Ce choix n’est pas complétement libre; le fait que l’effectif
d’échantillon est nécessairement un nombre entier impose certaines
NOTE - II est également nécessaire de rappeler que la qualité limite est contraintes. S’il est nécessaire d’avoir recours à un plan spécial, il est
la qualité qui, si on la présente à un contrôle, aurait une probabilité recommandé que celui-ci soit uniquement établi avec l’aide d’un statis-
d’acceptation de 10 %. Le risque réel assumé par le client dépend donc ticien compétent en matière de contrôle de la qualité.
60 3951 : 1989 (FI
Section trois: Mise en œuvre d’un plan d’échantillonnage par mesures
13 Opérations préliminaires indispensables
Ensuite, comparer la statistique de qualité (Q, et/ou Qi) avec la
constante d’acceptabilité k trouvée dans la table II-A pour le
Avant d’entreprendre un contrôle par mesures, il faut vérifier
contrôle normal. Si la statistique de qualité appropriée est supé-
rieure ou égale à la constante d’acceptabilité, le lot est accepta-
a) que la distribution peut être considérée comme normale
ble;; si elle est inférieure, il n’est pas acceptable.
et que la production est continue;
Ainsi, si une limite de spécification supérieure Ls est seule
b) s’il faut commencer par appliquer la méthode w) (ou
donnée, le lot est
«RH) ou, si l’écart-type est stable et connu, il faut dans ce
cas appliquer la méthode «o));
acceptable si 0, > k
c) que le niveau de contrôle à appliquer a été déterminé; si
aucun niveau n’a été prescrit, le niveau de contrôle II sera
non acceptable si 0, < k
utilisé;
d) que le NQA a été fixé et que celui-ci est un des NQA Ou, si une limite de spécification inférieure Li est seule donnée,
recommandés figurant dans la présente Norme internatio- le lot est
nale; sinon, les tables ne sont pas applicables;
acceptable si Qi > k
e) que, dans le cas de limites de spécification doubles, il a
été précisé si ces limites sont séparées ou combinées et, s’il
non acceptable si Qi < k
s’agit de limites séparées, que les NQA ont bien été fixés
pour chaque limite.
Si à la fois L, et Li sont données (k ayant des valeurs différentes
si les NQA sont différents pour la limite supérieure et la limite
inférieure), le lot est
14 Procédure normalisée pour l’application
de la méthode NS»
acceptable si à la fois Qi > ki et Q, > k,
14.1 Choix d’un plan
non acceptable si Qi < ki OU 0, < k,
La procédure pour choisir un plan est la suivante :
Exemple
a) Le niveau de contrôle étant donné (normalement ce
La température maximale de fonctionnement d’un certain
sera le niveau II) ainsi que l’effectif du lot, déterminer la
appareil est fixée à 60 OC. La production est contrôlée par lots
lettre-code a l’aide de la table I-A.
de 100 appareils. Sont prescrits : le niveau de contrôle II, le
b) Avec cette lettre-code et le NQA, utiliser la table II-A
contrôle normal avec un NQA de 2,5 %. D’après la table I-A, la
pour déterminer l’effectif d’échantillon n et la constante
lettre-code est F; en consultant la table Il-A, on constate qu’un
d’acceptabilité k.
échantillon d’effectif 10 est nécessaire et que la constante
d’acceptabilité k est de 1,41. Supposons que les mesures aient
c) Ayant prélevé au hasard un échantillon de cet effectif,
donné les résultats suivants : 53 OC; 57 OC; 49 OC; 58 OC;
mesurer le caractère x de chaque individu et calculer ensuite
54 OC; 58 OC; 56 OC; 55 OC; 50 OC. II est demandé de détermi-
la moyenne xde l’échantillon, ainsi que l’écart-type estimé s
ner si le critère d’acceptabilité est satisfait.
(voir annexe A). Si xest en dehors de la limite de spécifica-
tion, le lot est jugé inacceptable sans qu’il soit nécessaire de
Renseignements nécessaires Valeur obtenue
calculer s. Cependant, il peut être nécessaire de calculer s
pour la tenue du cahier d’enregistrement des résultats.
Effectif d’échantillon : n 10
14.2 Critères d’acceptabilité pour des limites de
Moyenne de l’échantillon X: C x/n
wg
spécification uniques ou doubles séparées
Écart-type S : 2/ C (Xi - Fi)*/(n - 1)
3,414
Si des limites de spécification uniques ou doubles séparées ont .
été fixées, calculer la statistique de qualité à l’aide des formules
(Voir A.l.2, annexe A.)
suivantes :
Limite de spécification (supérieure) : L,
-x
L
Q, = -
S
0, = (L, - x)/s
1,494
et/ou
Constante d’acceptabilité : k (voir table Il-A)
1,41
X- Li
Critère d’acceptabilité : comparer Q, avec k
Qi = - 1,494 > 1,41
S
Le lot satisfait au critére d’acceptabilité, il est par conséquent
selon le cas.
acceptable.
ISO 3951 : 1989 (FI
14.3 Méthode graphique pour une limite de de contrôle II, le contrôle normal, doit être appliqué avec un
NQA de Q,l % appliqué à la limite inférieure et un NQA de
spécification unique
2,5 % appliqué à la limite supérieure. On peut constater,
d’après la table I-A, que la lettre-code de l’effectif d’échantillon
Pour obtenir un critère graphique, tracer la ligne
est J; d’après la table I-B, que l’effectif d’échantillon est 35 pour
la méthode «SH; et d’après la table II-A, on constate que les
x= L, - ks (pour une limite supérieure) ou
constantes d’acceptabilité supérieure et inférieure sont
X= Li - ks (pour une limite inférieure)
k, = 1,57 et ki = 254, respectivement. Supposons que les
temps de retardement de l’échantillon soient les suivants:
en ayant soin d’utiliser du papier millimétré et de prendre X
comme axe vertical et s comme axe horizontal. Dans le cas où
6,95 604 6,@ W3 6,65
le contrôle s’applique à la limite de spécification supérieure, la
zone d’acceptation est la surface située au-dessous d
...
NORME
ISO
INTERNATIONALE
l
1989-09-15
Règles et tables d’échantillonnage pour les
contrôles par mesures des pourcentages de non
conformes
Sampling procedures and charts for inspection by variables for percent
noncon forming
Numéro de référence
ISO 3951 : 1989 (FI.
ISO3951:1989 (FI
Avant-propos
L’ISO (Organisation internationale de normalisation) est une fédération mondiale
d’organismes nationaux de normalisation (comités membres de I’ISO). L’élaboration
des Normes internationales est en général confiée aux comités techniques de I’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 I’ISO participent également aux travaux. L’ISO col-
labore étroitement avec la Commission électrotechnique internationale (CEI) en ce qui
concerne la normalisation électrotechnique.
Les projets de Normes internationales adoptés par les comités techniques sont soumis
aux comités membres pour approbation, avant leur acceptation comme Normes inter-
nationales par le Conseil de I’ISO. Les Normes internationales sont approuvées confor-
mément aux procédures de I’ISO qui requièrent l’approbation de 75 % au moins des
comités membres votants.
La Norme internationale ISO 3951 a été élaborée par le comité technique ISO/TC 69,
Applications des méthodes statistiques.
Cette deuxième édition annule et remplace la première édition (ISO 3951 : 1981) dont
elle constitue une révision technique.
Les principaux changements par rapport à la première édition sont les suivants:
a) une distinction a été faite entre l’écart-type maximal du processus (ETMP)
selon la méthode w)) et l’écart-type maximal de l’échantillon (ETME) selon la
méthode «s»;
b) toutes les courbes
d’acceptati on pour la méthode WJD ont été tronquées
aux
valeurs appropriées de l’écart-type maxima
I du processus;
c) la terminologie a été alignée sur celle de I’ISO 2859 et I’ISO 3534.
L’attention des utilisateurs est attirée sur le fait que toutes les Normes internationales
sont de temps en temps soumises à révision et que toute référence faite à une autre
Norme internationale dans le présent document implique qu’il s’agit, sauf indication
contraire, de la derniére édition.
0 ISO 1989
Droits de reproduction réservés. 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’éditeur.
Organisation internationale de normalisation
Case postale 56 l CH-1211 Genève 20 l Suisse
.
Imprimé en Suisse
ii
ISO 3951 : 1989 (F)
Sommaire
Page
Section un : Gén&alités. .
1 Objet et domaine d’application .
2 Références .
3 Définitions et symboles .
4 Niveau de qualité acceptable (NQA) .
.........
5 Règles pour la modification du contrôle normal, renforcé et réduit
6 Relations avec I’ISO 2859. .
7 Production non continue et courbes d’efficacité .
Section deux : Choix d’un plan d’échantillonnage .
8 Généralités sur le choix d’un plan .
9 Choix entre mesures et attributs .
10 Choix d’une méthode .
11 Choix du niveau de contrôle et du NQA .
12 Choix d’un plan d’échantillonnage .
Section trois : Mise en œuvre d’un plan d’échantillonnage par mesures . . 8
13 Opérations préliminaires indispensables .
............... 8
14 Procédure normalisée pour l’application de la méthode WH
Procédure normalisée pour l’application de la méthode ((ct)) . 13
Déroulement d’un contrôle continu . 16
17 Normalité et aberrants . 16
18 Enregistrement des résultats . 16
............... 17
19 Mise en œuvre des règles pour la modification du contrôle.
20 Suspension et reprise du contrôle. . 17
21 Courbes d’acceptation dans le cas du contrôle renforcé et du contrôle réduit
22 Passage à la méthode ((0)) .
. . .
III
Section quatre : Tables et graphiques . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
l
Tables
Lettres-code d’effectif d’échantillon et niveaux de contrôle . . .
I-A 19
Lettres-code d’effectif d’échantillon et
I-B
d’échantillon dans un contrôle normal . . . . 20
II-A Plans d’échantillonnage simples pour le contrôle normal
. .
(table générale) : méthode GSD . 21
II-B Plans d’échantillonnage simples pour le contrôle renforcé
(table générale) : méthode NS)) . . . 22
II-C Plans d’échantillonnage simples pour le contrôle réduit
(table générale) : méthode NS)) . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
d’échantillonnage simples le contrôle
Ill-A Plans pour
(table générale) : méthode WN . . . 24
Ill-B Plans d’échantillonnage simples pour le contrôle renforcé
générale) : méthode WN .
(table 25
III-C Plans d’échantillonnage simples pour le contrôle réduit
(table générale) : méthode WN . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
Valeurs de fs pour l’écart-type maximal de l’échantillon
IV-S
(ETME) : méthode w) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
IV-a Valeurs de fa pour l’écart-type maximal du processus
(ETMP) : méthode ((0)) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
\
V (V-B à V-P) Tables et courbes d’efficacité des plans d’échantillonnage
simples (lettres-code d’effectif d’échantillon B à P) . . . . . . . . . . 28
VI-A Correspondance entre le plan de contrôle renforcé
et le plan de contrôle normal . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
Correspondance entre le plan de contrôle réduit
VI-B
et le plan de contrôle normal . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
Graphiques
A Lettres-code d’effectifs d’échantillon des plans d’échantillonnage
normalisés tels que, pour des qualités spécifiées, les
probabilités d’acceptation soient de 95 % et 10 % . . . . . . . . . . . 43
s-D à s-P Courbes d’acceptation pour des limites de spécification
doubles combinees : méthode NS)) . . . . . . . . . . . . . . . . . . . . . . . . 44
a-c à a-P Courbes d’acceptation pour des limites de spécification
doubles combinées : méthode NON. . . . . 56
Annexes
A Calculdesetdea. . 69
................................................. 72
B Théorie statistique
C Plans d’échantillonnage pour la méthode (A). . 83
................................. 106
D Papier millimétré pour la méthode NS))
Bibliographie. 108
iv
NORME INTERNATIONALE
ISO 3951 : 1989 (FI
Règles et tables d’échantillonnage pour les contrôles
par mesures des pourcentages de non conformes
Section un : Généralités
1 Objet et domaine d’application d) lorsqu’un contrat ou une norme définit une limite
supérieure de spécification L,, une limite inférieure de
spécification Lit ou les deux; un produit est qualifié de non
1.1 Objet
conforme si la mesure de son caractère de qualité x satisfait
à l’une des inégalités suivantes:
1.1.1 La présente Norme internationale présente les plans et
x > L, . . .
(1)
règles d’échantillonnage pour les contrôles par mesures. Elle
complète I’ISO 2859. Les cahiers des charges, les contrats, les
X < Li . . .
(2)
instructions de contrôle ou autres textes peuvent se référer à la
soit x > L,, soitx < Li
. . . (3)
présente Norme internationale et à I’ISO 2859 lorsque l’autorité
responsable le prescrit, et les dispositions qu’elles contiennent
Les inégalités (1) et (2) répondent aux cas d’une limite unique
doivent être respectées. L’autorité responsable doit être dési-
de spécification et (3) au cas d’une limite double de spkifi-
gnée dans l’un de ces textes.
cation. Dans cette dernière situation, une distinction supplé-
mentaire est apportée entre les limites doubles séparées ou
1.1.2 Le but des méthodes présentées dans la présente
combinées, selon que le NQA est appliqué à chaque limite
Norme internationale est d’assurer que les lots de qualité accep-
séparément ou aux deux limites combinées (voir chapitre 4).
table ont une forte probabilité d’être acceptés et que la probabi-
lité de ne pas accepter les lots de qualité inférieure est aussi éle-
vée que possible.
2 Références
1.1.3 En accord avec I’ISO 2859, le pourcentage de produits ISO 2854, Interprétation statistique des données - Techniques
d’estimation et tests portant sur des moyennes et des
non conformes dans les lots est utilisé pour définir la qualité de
variances.
ces lots et des procédés de fabrication en question.
ISO 2859, Règles et tables d’échantillonnage pour les contrôles
1.2 Domaine d’application
par attributs,
La présente Norme internationale est principalement destinée à
ISO 3534, Statistique - Vocabulaire et symboles.
être utilisée dans les conditions suivantes :
ISO 5479, Tests de normalité. 1)
a) lorsque les règles de contrôle doivent être appliquées à des
séries continues de lots de produits individualises tous
I S 0 5725, Fidélité des me thodes d’essai - Dé termina tion de la
fournis par un seul producteur utilisant un seul procédé de
répé tabilité et de la reproductibilité d’une méthode d’essai nor-
fabrication. S’il y a différents producteurs, la présente Norme
malisée par essais in terlabora toires.
internationale doit être appliquée à chacun d’eux séparément;
b) lorsqu’un seul caractère de qualité, x, de ces pro-
duits est pris en considération, il doit être mesurable sur
3 Définitions et symboles
une échelle continue. Si plusieurs caractéres du produit
sont importants, la présente Norme internationale doit être
3.1 Définitions
appliquée à chacun d’eux séparément;
c) lorsque la fabrication est stable (sous contrôle stasti- Pour les besoins de la présente Norme internationale, les défini-
que) et que le caractère de qualité x est distribué suivant une tions données dans I’ISO 3534 et I’ISO 2859 ainsi que les défini- .
loi normale ou voisine d’une loi normale;
tions suivantes s’appliquent.
1) Actuellement au stade de projet.
ISO 3951 : 1989 (FI
3.1.1 contrôle par mesures (inspection by variables) : Con- 3.1.10 limite de spécif ication (specification limit) : Valeur
trole dans lequel on mesure un caractére quantitatif lie à chacun limite (minimale ou maximale) spécifiée pour un caractere
des individus d’une population ou d’un échantillon prélevé dans quantitatif.
cette population.
3.1.11 limite inférieure de spécification (LJ [lower specifi-
3.1.2 echantillonnage par mesures en vue d’acceptation
cation limit (L)] : Limite de spécification qui définit la limite infé-
(acceptance sampling by variables) : Procédure selon laquelle
rieure de reconnaissance de conformité d’une unité individuelle
un caractère spécifié est mesuré pour établir statistiquement
dans une fabrication ou un service.
I’acceptabilité du lot à partir des résultats obtenus sur les indivi-
dus d’un échantillon.
3.1.12 limite supérieure de spécification (L,) [Upper speci-
fication limit (U)] : Limite de spécification qui définit la limite
3.1.3 niveau de qualité acceptable (NQA) [acceptable
supérieure de reconnaissance de conformité d’une unité indivi-
quality level (AQL)] : Dans le cas d’une série continue de lots,
duelle dans une fabrication ou un service.
niveau de qualité qui, pour le contrôle par échantillonnage, est
la limite du pourcentage satisfaisant de non conformes de la
caractéristique moyenne de la qualité de production. (Voir cha-
3.1.13 limite unique de spécification (single specification
pitre 4.)
limit) : Terme utilisé lorsqu’il n’est spécifié qu’une seule limite.
3.1.4 qualite limite (limiting quality) : Pour un lot considéré
3.1.14 limites de specification doubles separées (sepa-
isolément, niveau de qualité qui, dans un contrôle par échantil-
rate double specification limits) : Terme utilisé lorsque sont
lonnage, correspond à une faible probabilité d’acceptation
spécifiées à la fois une limite supérieure et une limite inférieure
(dans la présente Norme internationale: 10 %). (Voir 12.1.)
et quand des NQA distincts sont appliqués à chacune des limi-
tes individuellement. (Voir 4.3.)
3.1.5 non-conformité (nonconformity) : Défaut de carac-
tére de qualité d’un produit ou service à ne pouvoir répondre à
3.1.15 limite de spécification double combinée (combi-
une exigence spécifiée, et dont l’évaluation ne dépend pas
ned double specification limit) : Terme utilisé lorsque sont spé-
essentiellement de la durée de temps.
cifiées à la fois la limite supérieure et la limite inférieure et quand
un NQA donné s’applique à l’ensemble des deux limites et cor-
Selon leur degré d’importance, les non-conformités sont géné-
respond au pourcentage total de non conformes. (Voir 4.3.)
ralement classées comme suit:
Classe A. Type de non-conformités ayant une incidence grave
3.1.16 constante d’acceptabilité (k) (acceptability cons-
sur le produit ou le service. Dans l’échantillonnage en vue
tant) : Constante dépendant de la valeur assignée au niveau de
d’acceptation, on fixera pour de tels types de non-conformités
qualité acceptable et de l’effectif de l’échantillon. (Voir 14.2 et
une valeur très faible pour le NQA.
15.2, éventuellement chapitre C.5 de l’annexe C.)
Classe B. Type de non-conformités ayant une incidence moins
3.1.17 statistique de qualité (0) (quality statistic) : Fonc-
importante et pour lesquelles sera fixée une valeur du NQA
supérieure à celles de la classe A et inférieure à celles de la tion de la limite de spécification, de la moyenne de l’échantillon
et de l’écart-type de l’échantillon. La décision sur le lot est prise
classe C, si cette dernière existe, et ainsi de suite.
en comparant 0 avec la constante d’acceptabilité k. (Voir 14.2
Le nombre des classes et l’affectation d’une non-conformité et 15.2, éventuellement chapitre C.5 de l’annexe C.)
dans une classe donnée doivent être adaptés aux exigences de
qualité de la situation particulière considérée.
3.1.18 statistique de qualité correspondant à la limite
inférieure (Qi) [lower quality statistic (QL)l : Fonction de la
3.1.6 unité non conforme (nonconforming unit) : Unité
de limite inférieure de spécification, de la moyenne de l’échantillon
produit ou de service contena nt au moins une non-conformi té. et de l’écart-type de l’échantillon. La décision sur le lot est prise
en comparant Qi avec la constante d’acceptabilité k. (Voir 14.2
et 15.2 ou chapitre C.5 en annexe C).
3.1.7 méthode (w) (‘Y method) : Méthode permettant de
décider si un lot est acceptable à partir de l’écart-type de
l’échantillon. (Voir chapitre 14.)
3.1.19 statistique de qualité correspondant à la limite
supérieure (Q,) [Upper quality statistic (QU)I : Fonction de la
limite supérieure de spécification, de la moyenne de I’échantil-
3.1.8 méthode (WH (“0” méthod) : Méthode permettant de
Ion et de l’écart-type de l’échantillon. La décision sur le lot est
décider si un lot est acceptable à partir de la connaissance que
prise en comparant Q, avec la constante d’acceptabilité k.
l’on a de l’écart-type du processus. (Voir chapitre 14.)
(Voir 14.2 et 15.2 ou chapitre C.5 en annexe C.)
3.1.9 methode c(R)) (‘Y?” method) : Méthode permettant de
décider si un lot est acceptable en adoptant une estimation indi-
3.120 écart-type maximal de l’échantillon (ETME) [maxi-
recte de l’écart-type du processus à partir de l’étendue
mum sample standard deviation (MSSD)] : Dans des conditions
moyenne des mesures effectuées par sous-groupes d’individus
données, le plus grand écart-type de l’échantillon acceptable.
d’un échantillon. (Voir annexe C.)
(Voir 14.6 et B.8.3 dans l’annexe B.)
ISO 3951 : 1969 (FI
3.1.21 écart-type maximal du processus (ETMP) [maxi- d’acceptation maximale.
Valeur
xs
b
mum process standard deviation (MPSD)]: Dans des condi-
tions données, le plus grand écart-type du processus accepta- Moyenne du processus
ru
ble. (Voir 15.3 et B.5.2 dans l’annexe B.)
processus (a*, carré de l’écart-type, est
ct Écart-type du
appelé variante).
3.122 règles de modification du contrôle (switching
rules) : Instructions, contenues dans une procédure d’échantil-
C «Somme de)) (par exemple, C x = somme de tous
lonnage, passage d’un plan d’échantillonnage à un autre, à par-
les x).
tir d’un compte-rendu prouvé de la qualité. (Voir chapitre 19.)
n
Somme de tous les Xi quand i prend toutes les
xi
c
3.2 Symboles valeurs de 1 à n.
i=
> «Supérieur à» (par exemple, a > b signifie que a est
Les symboles employés sont les suivants:
supérieur à b).
& Facteur, donné dans la table IV-s, qui relie l’écart-type
«Supérieur ou égal à» (par exemple, a > b signifie que
maximal de I ‘échantillon à la différence entre L, et Li.
a est supérieur ou égal à 6).
fa Facteur, donné dans la table IV-a, qui relie l’écart-type
«Inférieur à» (par exemple, a < b signifie que a est
maximal du processus à la différence entre
L, et Lia
inférieur à 6).
k Constante d’acceptabilité quand on utilise la méthode
«Inférieur ou égal à» (par exemple, a < b signifie que a
as)), mn) ou (db.
est inférieur ou égal à b).
K Constante d’acceptabilité lorsque ,u et 0 sont connus.
3.3 Bibliographie
)), en indice à une
Li Limite inférieure de spécification ((4
Une bibliographie des docu ments uti Iisés pour l’élaboration de
variable, signifie valeur de cette variable à Li).
la présente Norme internati onale est donnée dans l’annexe D.
L, Limite supérieure de spécif ication (« :s)), en indice à une
variable, signifie valeu r de cette variable à L,).
4 Niveau de qualité acceptable (NQA)
d’échantillon (nombre d’unités dans un échan-
n Effectif
4.1 Définition
tillon).
Dans le cas d’une série continue de lots, le NQA est le niveau de
N Effectif du lot (nombre d’unités dans un lot).
qualité qui, pour le contrôle par échantillonnage, est la limite du
pourcentage satisfaisant de non conformes de la caractéristi-
pa Probabilité d’acceptation.
que moyenne de la qualité de production.
0 Statistique de qualité.
4.2 Utilisation
Qi Statistique de qualité correspondant à la limite infé-
rieure.
Le NQA est utilisé en même temps que la lettre-code d’effectif
d’échantillon pour repérer les plans d’échantillonnage décrits
Q, Statistique de qualité correspondant à la limite supé-
dans la présente Norme internationale.
rieure.
4.3 Spécification du NQA
s Écart-type d’un échantillon estimation de l’écart-type
processus)
Le (les) NQA utilisé(s) doivent être fixé(s) dans le contrat de
spécification du produit ou par l’autorité responsable. Si des
n
limites de spécification supérieure et inférieure sont données,
(Xi - X)*
des NQA séparés peuvent être attribués aux limites individuel-
c
i= 1 les, qui sont alors considérées comme des limites de spécifica-
=
S
tion doubles séparées. Alternativement, un seul NQA peut être
n-l
/
spécifié pour le pourcentage total de non conformes s’appli-
quant à la fois à la limite supérieure et à la limite inférieure; on a
(Voir aussi annexe A.)
alors affaire à une limite de spécification double combinée.
x Valeur mesurée d’un caractère dans l’échantillon.
4.4 NQA recommandés
X Moyen ne arithmétique de x pour un échantillon de n
individus.
Onze NQA donnés dans la présente Norme internationale, dont
les valeurs s’échelonnent de 0,IO % à 10 % de non conformes,
-
Valeur d’acceptation minimale.
sont dits «NQA recommandés». Si, pour n’importe quel produit
xi
60 3951 : 1989 (FI
5.6 Des détails relatifs aux règles de modification du contrôle
ou service, on spécifie une valeur de NQA autre qu’une autre
valeur recommandée, la présente Norme internationale n’est sont donnés dans les chapitres 19 et 20.
alors pas applicable. (Voir 12.2.)
Deux NQA supplémentaires, 0,085 % et 15 %, sont donnés
6 Relations avec I’ISO 2859
dans le but de compléter la gamme des plans que nécessite
l’application des règles de modification du contrôle. (Voir cha-
pitres 19 et 21.) Les plans et les courbes désignés par un NQA 6.1 Similitudes
de 0,065 % ou de 15 % ne,seront utilisés que lorsque le NQA,
dans le contrôle normal, est de 0,lO % ou de 10 % respective-
a) La présente Norme internationale complète I’ISO 2859;
ment, et que les règles de modification du contrôle sont invo-
les deux documents ont une philosophie commune et,
quées.
autant que possible, leurs règles et vocabulaire sont identi-
ques.
4.5 Remarque
b) Tous deux utilisent le NQA pour indexer les plans
d’échantillonnage, et les valeurs recommandées utilisées
D’après la définition du NQA donnée en 4.1, il s’ensuit que la
dans ce document sont identiques à celles qui sont données
protection désirée ne peut être obtenue que si une série conti-
dans I’ISO 2859 pour la même étendue de valeurs k’est-à-
nue de lots est présentée au contrôle.
dire de 0,l % à 10 %).
c) Dans les deux Normes internationales, l’effectif du lot et
4.6 Limitation
le niveau de contrôle (le niveau II étant recommandé à
défaut d’autres instructions) déterminent une lettre-code.
La fixation d’un NQA n’implique pas pour le fournisseur le droit
Des tables générales donnent ensuite, en fonction de la
de livrer sciemment des individus non conformes.
lettre-code et du NQA fixé, l’effectif de l’échantillon à préle-
ver et le critère d’acceptabilité suivant la méthode adoptée
(w), GO)), ou éventuellement, «R»I. Des tables distinctes
5 Règles pour la modification du contrôle
sont données pour le contrôle normal, le contrôle renforcé
et le contrôle réduit.
normal, renforcé et réduit
d) Les règles pour la modification du contrôle sont essen-
5.1 Afin de décourager le producteur d’opérer avec un pour-
tiellement les mêmes.
centage de non conformes qui dépasse celui préconisé par le
e) La classification des non-conformités en fonction du
NQA, la présente Norme internationale prescrit de passer au
degré de leur importance en classe A, classe B, etc. n’est
contrôle renforcé dans le cas où les résultats du contrôle font
pas modifiée.
présumer que la qualité moyenne de production est inférieure à
celle qui est caractérisée par le NQA, et de suspendre entière-
ment le contrôle par échantillonnage dans le cas où le contrôle
6.2 Différences
renforcé n’encourage pas à temps le producteur à améliorer son
processus de fabrication.
a) Détermination de I’acceptabilité. L’acceptabilité,
dans un plan d’échantillonnage par attributs de I’ISO 2859,
5.2 Le contrôle renforcé et la règle d’interruption font partie est déterminée par le nombre d’unités non conformes trou-
intégrante de la présente Norme internationale et, par consé- vées dans l’échantillon; le critère d’acceptabilité dans le con-
quent, constituent des procédures d’application obligatoire
trôle par mesures est basé sur les estimations de la tendance
pour maintenir la protection obtenue pour un NQA donné. centrale et de la variabilité de la distribution des mesures
dans le lot, en relation avec les limites de spécification. Ces
notions sont traduites en termes de moyenne et d’écart-
5.3 La présente Norme internationale prévoit aussi la possibi-
type. Dans la présente Norme internationale, deux métho-
lité de passer à un contrôle réduit si les résultats du contrôle
la méthode w) que l’on applique
des sont présentées :
montrent que le pourcentage des non conformes dans la qua-
dans le cas où l’écart-type du processus 0 est inconnu et la
lité moyenne de fabrication est stable et meilleure que le NQA.
méthode ((0)) que l’on applique dans le cas où 0 est consi-
Cette méthode est cependant facultative (a la discrétion de
déré comme étant connu. Une troisième méthode, appelée
l’autorité responsable) l
méthode c(R)), est donnée dans l’annexe C. Dans le cas
d’une limite unique de spécification ou de deux limites sépa-
rées, l’acceptabilité peut être calculée à partir d’une formule
5.4 S’il existe des preuves suffisantes d’après les cartes de
(voir 14.2 et 15.21, mais elle est établie plus facilement par
contrôle (voir 18.1) que la variabilité est sous contrôle, il peut
une méthode graphique (voir 14.3). Dans le cas d’une limite
être envisagé de passer à la méthode ((a)). Si cette méthode est
double combinée, la présente Norme internationale prévoit
effectivement avantageuse, la valeur à adopter pour a est la
une méthode graphique (voir 14.4 et 15.3).
valeur stable de s.
b) Normalité. Dans I’ISO 2859, aucune exigence relative
5.5 S’il a été nécessaire d’interrompre le contrôle par échan- à la distribution des caractères n’est stipulée; par contre
dans la présente Norme internationale, il est nécessaire,
tillonnage, le contrôle renforcé ne peut pas être repris avant
que des actions aient été prises par le fabricant pour améliorer pour l’application correcte du plan, que les mesures soient
distribuées suivant une loi normale ou voisine de la normale.
la qualité du produit.
1s0 3951 : 1989 (FI
c) Courbes d’efficacité (OC curves). Bien que l’on
7 Production non continue et courbes
puisse concevoir un plan d’échantillonnage par mesures
d’efficacité
dont la courbe d’efficacité corresponde étroitement à celle
d’un plan par attributs, il ne serait pas possible d’établir,
7.1 Production non continue
dans la présente Norme internationale, toutes les courbes
d’efficacité correspondant à celles de I’ISO 2859 (indexées
a) Les méthodes d’échantillonnage contenues dans la pré-
selon la même lettre-code et le même NQA) sans que
sente Norme internationale ne sont pas destinées à être
l’effectif d’échantillon n’augmente avec le NQA pour une
appliquées dans des conditions différentes de celles qui sont
lettre-code d’effectif d’échantillon donnée. Pour la méthode
spécifiées en 1.2, par exemple pour un lot isolé ou un nom-
(~1, on a gardé le même effectif d’échantillon pour un effec-
bre limité de lots, pour lesquels le contrôle renforcé et les
tif de lot donné, dans toute la gamme des NQA; pour la
règles d’interruption ne peuvent pas être appliqués.
méthode w)), cette contrainte a été levée, afin d’associer
b) Dans ces conditions, le concept de NQA n’est plus petti-
aussi étroitement que possible la courbe d’efficacité de la
nent, dans la mesure où le consommateur ne se soucie que de
méthode w) et celle de la méthode HO)), à la fois au NQA et
la qualité du nombre limité des lots présentés, cessant d’exer-
à la qualité limite.
cer des contrôles sur la qualité du processus de fabrication.
Le NQA indique cependant une qualité présentée qui a une
d) Probabilité d’acceptation pour un NQA. La proba-
grande probabilité d’acceptation et peut donc servir de repé-
bilité qu’un lot, dont la qualité est égale à celle qui est carac-
rage d’un plan d’échantillonnage.
térisée par le NQA, soit accepté augmente avec l’effectif de
l’échantillon et suit une loi analogue mais non identique à
7.2 Courbes d’efficacité
celle qui est utilisée dans I’ISO 2859.
a) Le degré de protection du client obtenu pour chacun
des différents plans d’échantillonnage de cette norme peut
e) Effectifs d’échantillon. Les effectifs d’échantillon
néanmoins être évalué à partir des courbes d’efficacité don-
pour le contrôle par mesures correspondant à des lettres-
nées sur les graphiques V-B à V-P et dans les tables V-B-l à
code données sont généralement plus petits que les effec-
V-P-l, qui doivent être consultés pour le choix d’un plan
tifs d’échantillon pour le contrôle par attributs correspon-
d’échantillonnage.
dant aux mêmes lettres-code.
b) Ces courbes sont destinées au contrôle normal pour
l’application de la méthode w) avec une seule limite de spé-
f) Plans d’échantillonnage doubles. Aucun plan
cification, mais elles fournissent une bonne approximation
d’échantillonnage double n’est donné dans la présente
dans le cas d’une limite de spécification double combinée.
Norme internationale.
c) Ces courbes fournissent également une bonne approxi-
mation des courbes d’efficacité pour les plans d’échantillon-
g) Limite de la qualité moyenne après contrôle
nage de la méthode ((0)) (et de la méthode ((R))I repérés par
(AOQL). Lors d’essai destructifs ou coûteux, dans lesquels
la même lettre-code et le même NQA, sauf si l’effectif de
le contrôle et le remplacement à 100 % des lots non accep-
l’échantillon est petit.
tés n’est pas possible, le concept d’AOQL ne peut pas être
appliqué. Dans la mesure où des plans de contrôle par d) Les courbes d’efficacité séparées ne sont pas présen-
mesures sont généralement appliqués dans de tels cas, les tées pour le contrôle renforcé ou réduit, mais les courbes
tables correspondantes d’AOQL ne sont pas présentées peuvent être trouvées parmi celles données pour le contrôle
dans la présente Norme internationale. normal (voir tables VI-A et VI-B).
ISO 3951 : 1989 (FI
.
Section deux: Choix d’un plan d’échantillonnage
la distribution des mesures est normale. S ‘il existe un doute,
8 Généralités sur le choix d’un plan
être consultée
l’autorité responsable doit
Le choix du plan par mesures le mieux approprié, s’il en existe
un, requiert de l’expérience, du jugement et une assez bonne
NOTES
connaissance a la fois de la statistique et du produit à contrôler.
1 Des tests de normalité figurent dans la section deux de I’ISO 2854
La présente section vise à suggérer, aux responsables chargés
qui donne des exemples de méthodes graphiques pouvant être utilisées
de déterminer les plans d’échantillonnage, les considérations
pour vérifier qu’une distribution de données est suffisamment proche
qu’ils doivent avoir présentes à l’esprit quand ils décident si un
de la normale pour justifier l’utilisation de plans d’échantillonnage par
plan par mesures conviendrait bien et quels sont les choix à
mesures.
faire pour sélectionner un plan normalisé approprié.
2 Une documentation plus précise concernant les tests de normalité
est donnée dans I’ISO 5479.
9 Choix entre mesures et attributs
10 Choix d’une méthode
La première question qui se pose est de décider s’il est préféra-
ble de faire un contrôle par mesures plutôt qu’un contrôle par
Si l’on a décidé d’appliquer un contrôle par mesures, la pre-
attributs. II est recommandé de prendre en considération les
mière question qui se pose est de savoir quelle est la méthode à
points suivants :
employer : la méthode w) ou la méthode ((CT)) (éventuellement,
la méthode «R»).
a) Au point de vue économique, on doit comparer le coût
total d’un contrôle relativement simple d’un plus grand
La méthode ((0)) est la plus économique au point de vue de
nombre d’individus à l’aide d’une méthode par attributs à
l’effectif de l’échantillon mais, avant que cette méthode puisse
celui d’une procédure généralement plus élaborée, exigée
être employée, la valeur de ~0)) doit être déterminée.
par une méthode par mesures qui revient généralement plus
cher en temps et en argent dépensés par élément contrôlé.
Si l’on considère l’effectif d’échantillon, la méthode w) pré-
b) Au point de vue de l’information obtenue, l’avantage
sente un léger avantage sur la méthode «R» mais la détermina-
revient au contrôle par mesures pour lequel on obtient un
tion de s entraîne plus de calculs; l’importance et la difficulté de
renseignement plus précis sur le niveau de qualité du pro-
ceux-ci est plus apparente que réelle, particulièrement si l’on
duit : on sera prévenu plus tôt si la qualité dérive.
dispose d’un calculateur électronique. Les méthodes de calcul
de s sont données dans l’annexe A.
c) Une méthode par attributs peut être plus aisément com-
prise et acceptée; par exemple, il peut être difficile d’admet-
La méthode «R)) (donnée dans l’annexe C) est simple à calcu-
tre d’emblée que, lorsqu’on effectue un contrôle par mesu-
ler, mais elle exige un effectif d’échantillon un peu plus impor-
res, un lot puisse être rejeté à partir de mesures faites sur un
tant pour le même NQA. En outre, pour des échantillons
échantillon ne contenant pas de non conformes. (Voir
l’exemple en 14.6.) d’effectif 10 ou plus, I’acceptabilité d’un lot peut dépendre de la
manière dont est divisé l’échantillon en sous-groupes, ce qui
d) Une comparaison entre les effectifs d’échantillon exigés
constitue une propriété indésirable de la méthode.
pour obtenir le même NQA, en utilisant des plans normali-
sés du contrôle par attributs (c’est-à-dire I’ISO 2859) et les
Initialement, on devra commencer par la méthode w) (ou,
plans normalisés contenus dans la présente Norme interna-
éventuellement, la méthode c
tionale figure dans la table I-B. On remarquera que les
sante, les règles normalisées pour la modification du contrôle
échantillons les plus petits sont obtenus par la méthode
permettront, à l’autorité responsable, d’introduire un contrôle
w)), utilisée quand l’écart-type du processus est connu.
réduit et de prendre un effectif d’échantillon plus petit.
e) Le contrôle par mesures convient particulièrement bien
quand il est conjugué avec l’utilisation de cartes de contrôle
La question devient alors : si la variabilité est sous contrôle et si
par mesures.
les lots continuent d’être acceptés, serait-il plus économique de
passer à la méthode «a»?
f) L’échantillonnage par mesures offre un avantage consi-
dérable quand la méthode de contrôle est onéreuse, par
L’effectif de l’échantillon sera généralement moindre et le cri-
exemple, dans le cas d’un essai destructif.
tère d’acceptabilité sera plus simple. (Voir 15.2 et 15.3.) D’autre
g) Une méthode par mesures convient moins bien lorsque
part, il restera nécessaire de calculer s afin de pouvoir en enre-
s’accroît le nombre de caractères mesurés sur un même
gistrer les résultats et de tenir à jour les cartes de contrôle. (Voir
individu, puisqu’il y a lieu d’examiner chaque caractère
chapitre 18. )
séparément. On peut avoir avantage à appliquer le contrôle
par attributs à la majorité des caractères et le contrôle par
mesures à une ou deux des exigences les plus importantes,
11 Choix du niveau de contrôle et du NQA
par exemple : les essais de résistance à la charge, les exi-
gences de sécurité et de fiabilité.
Dans les plans d’échantillonnage normalisés, le niveau de con-
h) L’emploi de la présente Norme internationale n’est trôle, conjugué avec l’effectif des lots et avec le NQA, déter-
applicable que lorsqu’on a de bonnes raisons de croire que
mine l’effectif d’échantillon à prélever et régit la sévérité du
ISO 3951 : 1989 (F)
des marchandises de qualité inférieure
contrôle. La courbe d’efficacité appropriée, donnée dans l’une également de la probabilité que
soient préSentées au contrôle.
des tables V-B à V-P, montre l’importance du risque inhérent
au ptan.
Cependant, si dans certaines circonstances, la qualité limite a
Le choix du niveau de contrôle et du NQA est fonction de plu- une priorité plus grande que l’effectif d’échantillon (par exem-
ple, quand la production comporte seulement un nombre limité
sieurs facteurs mais dépend principalement d’un compromis
des lots), il est possible de choisir un plan de la présente Norme
équilibré entre le coût total du contrôle et les conséquences de
la mise en service de produits non conformes. internationale bien adapté, en employant le graphique A.
L’intersection de la ligne verticale passant par la valeur choisie
La procédure habituelle consiste à appliquer un contrôle au de la qualité limite et de la ligne horizontale passant par la qua-
niveau II jusqu’à ce que des circonstances spéciales indiquent lité acceptable pour laquelle la probabilité d’acceptation est de
qu’un autre niveau est préférable. 95 % (approximativement égale au NQA) se situera sur ou
au-dessous d’une ligne inclinée indexée par la lettre-code
d’effectif d’échantillon du plan normalisé répondant aux condi-
12 Choix d’un plan d’échantillonnage tions spécifiées. Cela peut être vérifié en examinant la courbe
d’efficacité donnée dans la table V pour cette lettre-code
d’effectif d’échantillon et ce NQA.
12.1 Plans normalisés
Si ces lignes se coupent en un point situé au-dessus de la ligne
La procédure normalisée ne peut être utilisée que lorsque la
marquée «PH (voir graphique A), cela implique que, par exem-
production des lots est continue.
ple, un échantillon de plus de 200 individus serait nécessaire
pour la méthode s, et donc que les conditions spécifiées ne
Cette procédure, avec ses étapes semi-automatiques depuis
peuvent pas être remplies au moyen des plans normalisés.
l’effectif du lot jusqu’à l’effectif d’échantillon, en utilisant le
niveau de contrôle II et en commentant par la méthode MB,
s’est révélée, en pratique, fournir des plans d’échantillonnage
12.2 Plans spéciaux
fonctionnant de facon acceptable, mais elle suppose que
l’ordre de priorité soit d’abord le NQA, ensuite l’effectif
Si les plans normalisés ne peuvent pas être retenus, il est
d’échantillon et en dernier lieu la qualité limite.
nécessaire de concevoir un plan spécial. Le choix consiste alors
à décider quelle combinaison de NQA, qualité limite et effectif
La valeur de ce système est due au fait que le client est protégé
d’échantillon, convient le mieux, en se rappelant que ces élé-
par les règles de passage d’un niveau de contrôle à un autre
ments ne sont pas indépendants car il suffit d’en choisir deux
(voir chapitre 19) qui augmentent rapidement la sévérité du
pour que le troisième en découle.
contrôle et finalement y mettent fin, si la qualité de la fabrica-
tion est plus mauvaise que celle caractérisée par le NQA.
NOTE - Ce choix n’est pas complétement libre; le fait que l’effectif
d’échantillon est nécessairement un nombre entier impose certaines
NOTE - II est également nécessaire de rappeler que la qualité limite est contraintes. S’il est nécessaire d’avoir recours à un plan spécial, il est
la qualité qui, si on la présente à un contrôle, aurait une probabilité recommandé que celui-ci soit uniquement établi avec l’aide d’un statis-
d’acceptation de 10 %. Le risque réel assumé par le client dépend donc ticien compétent en matière de contrôle de la qualité.
60 3951 : 1989 (FI
Section trois: Mise en œuvre d’un plan d’échantillonnage par mesures
13 Opérations préliminaires indispensables
Ensuite, comparer la statistique de qualité (Q, et/ou Qi) avec la
constante d’acceptabilité k trouvée dans la table II-A pour le
Avant d’entreprendre un contrôle par mesures, il faut vérifier
contrôle normal. Si la statistique de qualité appropriée est supé-
rieure ou égale à la constante d’acceptabilité, le lot est accepta-
a) que la distribution peut être considérée comme normale
ble;; si elle est inférieure, il n’est pas acceptable.
et que la production est continue;
Ainsi, si une limite de spécification supérieure Ls est seule
b) s’il faut commencer par appliquer la méthode w) (ou
donnée, le lot est
«RH) ou, si l’écart-type est stable et connu, il faut dans ce
cas appliquer la méthode «o));
acceptable si 0, > k
c) que le niveau de contrôle à appliquer a été déterminé; si
aucun niveau n’a été prescrit, le niveau de contrôle II sera
non acceptable si 0, < k
utilisé;
d) que le NQA a été fixé et que celui-ci est un des NQA Ou, si une limite de spécification inférieure Li est seule donnée,
recommandés figurant dans la présente Norme internatio- le lot est
nale; sinon, les tables ne sont pas applicables;
acceptable si Qi > k
e) que, dans le cas de limites de spécification doubles, il a
été précisé si ces limites sont séparées ou combinées et, s’il
non acceptable si Qi < k
s’agit de limites séparées, que les NQA ont bien été fixés
pour chaque limite.
Si à la fois L, et Li sont données (k ayant des valeurs différentes
si les NQA sont différents pour la limite supérieure et la limite
inférieure), le lot est
14 Procédure normalisée pour l’application
de la méthode NS»
acceptable si à la fois Qi > ki et Q, > k,
14.1 Choix d’un plan
non acceptable si Qi < ki OU 0, < k,
La procédure pour choisir un plan est la suivante :
Exemple
a) Le niveau de contrôle étant donné (normalement ce
La température maximale de fonctionnement d’un certain
sera le niveau II) ainsi que l’effectif du lot, déterminer la
appareil est fixée à 60 OC. La production est contrôlée par lots
lettre-code a l’aide de la table I-A.
de 100 appareils. Sont prescrits : le niveau de contrôle II, le
b) Avec cette lettre-code et le NQA, utiliser la table II-A
contrôle normal avec un NQA de 2,5 %. D’après la table I-A, la
pour déterminer l’effectif d’échantillon n et la constante
lettre-code est F; en consultant la table Il-A, on constate qu’un
d’acceptabilité k.
échantillon d’effectif 10 est nécessaire et que la constante
d’acceptabilité k est de 1,41. Supposons que les mesures aient
c) Ayant prélevé au hasard un échantillon de cet effectif,
donné les résultats suivants : 53 OC; 57 OC; 49 OC; 58 OC;
mesurer le caractère x de chaque individu et calculer ensuite
54 OC; 58 OC; 56 OC; 55 OC; 50 OC. II est demandé de détermi-
la moyenne xde l’échantillon, ainsi que l’écart-type estimé s
ner si le critère d’acceptabilité est satisfait.
(voir annexe A). Si xest en dehors de la limite de spécifica-
tion, le lot est jugé inacceptable sans qu’il soit nécessaire de
Renseignements nécessaires Valeur obtenue
calculer s. Cependant, il peut être nécessaire de calculer s
pour la tenue du cahier d’enregistrement des résultats.
Effectif d’échantillon : n 10
14.2 Critères d’acceptabilité pour des limites de
Moyenne de l’échantillon X: C x/n
wg
spécification uniques ou doubles séparées
kart-type s : J C (xi - X)2/(n - 1)
3,414
Si des limites de spécification uniques ou doubles séparées ont .
été fixées, calculer la statistique de qualité à l’aide des formules
(Voir A.l.2, annexe A.)
suivantes :
Limite de spécification (supérieure) : L,
-x
L
Q, = -
S
0, = (L, - x)/s
1,494
et/ou
Constante d’acceptabilité : k (voir table Il-A)
1,41
X- Li
Critère d’acceptabilité : comparer Q, avec k
Qi = - 1,494 > 1,41
S
Le lot satisfait au critére d’acceptabilité, il est par conséquent
selon le cas.
acceptable.
ISO 3951 : 1989 (FI
14.3 Méthode graphique pour une limite de de contrôle II, le contrôle normal, doit être appliqué avec un
NQA de Q,l % appliqué à la limite inférieure et un NQA de
spécification unique
2,5 % appliqué à la limite supérieure. On peut constater,
d’après la table I-A, que la lettre-code de l’effectif d’échantillon
Pour obtenir un critère graphique, tracer la ligne
est J; d’après la table I-B, que l’effectif d’échantillon est 35 pour
la méthode «SH; et d’après la table II-A, on constate que les
x= L, - ks (pour une limite supérieure) ou
constantes d’acceptabilité supérieure et inférieure sont
X= Li - ks (pour une limite inférieure)
k, = 1,57 et ki = 254, respectivement. Supposons que les
temps de retardement de l’échantillon soient les suivants:
en ayant soin d’utiliser du papier millimétré et de prendre X
comme axe vertical et s comme axe horizontal. Dans le cas où
6,95 6,65
le contrôle s’applique à la limite de spécification supérieure, la 604 6,@ W3
zone d’acceptation est la surface située au-dessous de la ligne.
64 644 6,15
634 604
Dans le cas où l’on s’intéresse à la limite de spécification infé-
7,15 6,70 6,51
64 659
rieure, la zone d’acceptation est la surface située au-dessus de
la ligne. En se référant aux valeurs de s et de xc
...
















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...